European Journal of Parapsychology v23-2 2008

European Journal of Parapsychology
Volume 23.2
2008
European Journal of Parapsychology © 2008 European Journal of Parapsychology
Volume 23.2, pages 131-146 ISSN: 0168-7263
Correlations between the EEGs of two spatially

separated subjects − a replication study
Wolfgang Ambach
Institute for Frontier Areas of Psychology and Mental Health, Freiburg, Germany
Abstract
Previous studies have reported correlations between the EEGs of

two spatially separated subjects. These correlations could not be
explained by means of conventional physiological or physical
mechanisms. Results by Wackermann (2003, 2004; Wackermann et
al, 2004) suggest such unexplained correlations, with the direction
(increase or decrease of EEG power) as well as the localization
(electrode position) of the supposed effects varying between studies.
The present replication study, performed in another laboratory, yet
other experimental changes kept at a minimum, aimed at clarifying
this effect. Seventeen pairs of subjects were examined. One subject
of each pair was visually stimulated by one-second periods of
checkerboard reversals while the other sat in a dimly lit separate
room without stimulation. The latter subject’s EEG power in the
paired subject’s stimulation and interstimulus periods was
compared; separate analyses were performed for two different time
windows. In this study, the previously used mode of EEG segment
sampling turned out to systematically overestimate effects.
Therefore, the non-parametric statistical method was modified.
Also, the results for the nineteen EEG channels were corrected for
their degrees of freedom using a Monte-Carlo simulation. Results
did not show a significant deviation from random expectation; the
EEG power of the non-stimulated subject was not linked to the
stimulation of the other subject.
Correspondence details: Wolfgang Ambach, Institut für Grenzgebiete der Psychologie und
Psychohygiene, Wilhelmstr. 3A, D-79098 Freiburg, Germany. Email: ambach@igpp.de
131
EEG correlations between separated subjects
Introduction
Since 1965, a series of studies have reported correlations between

the EEGs of spatially separated subjects. In these experiments, pairs of
subjects were prevented from interacting via known physiological or
physical mechanisms; consequently, the findings have often been
discussed in terms of direct brain-to-brain communication (Charman,
2006).
Grinberg-Zylberbaum et al. (1994) coined the term 'transferred
potentials' for the observation of correlations between the EEGs of two
spatially separated subjects of whom only one was visually stimulated.
They found statistical correlations between certain parts of the two
time series of the averaged event-related EEG segments obtained from
both subjects. From there on, event-related EEG correlations between
distant human subjects were mostly investigated on the basis of
averaged EEG segments. A critical analysis of the Grinberg-
Zylberbaum et al. (1994) study has been provided by May et al. (2001).
They provided evidence for the claim that violations of the underlying
assumptions concerning hypothesis testing had led to an
overestimation of the effects in this study. Several subsequent
experiments searching for EEG correlations which could not be
explained by a known mechanism, yielded positive results (e.g.
Fenwick et al., 1998), while other replication studies (e.g. Sabell et al.,
2001) failed to confirm the effect. Recently, correlations between brain
electrical activities of distant human brains were found by Standish
(2004) and Radin (2003, 2004).
Two studies by Wackermann et al. (2003, 2004) aimed at
examining if correlations in the EEG between a visually stimulated
subject and a second, non-stimulated subject were replicable when
communication between both participants was prevented by spatial
separation of subjects in shielded rooms. One subject ('subject A') was
visually stimulated with repeated series of checkerboard reversals
while the other subject ('subject B'), sitting in a dimly lit room, was not
stimulated. In the control condition, stimulation of the first subject was
prevented by covering the monitor with an opaque shield. From the
first to the second study, several methodological refinements had been
introduced. In both studies, the authors found significant fluctuations
of the EEG power of subject B according to the stimulation or
interstimulus periods of subject A. Remarkably, the effect of subject A’s
132
Ambach
stimulation on the EEG power of the non-stimulated subject B varied

between studies in direction (increase, decrease) as well as location
(electrode position). Furthermore, deviations from random expectation
were not only observed in the experimental condition (stimulation),
but partially also in the control condition (no stimulation). Thus,
significant results were generally replicated, but not the specific
finding of the direction of the effect or the location of its maximum;
results of the repeated studies were inconsistent but still striking.
Our replication study aimed at validating the existence of the
effect and, if possible, shedding light upon its specifications. In relation
to the preceding studies, it was conducted in a different laboratory and
guided by a different experimenter. All other experimental conditions
were kept unchanged as far as possible. The statistical methods applied
in the precedent studies were reconsidered in detail.
Method
Subjects
Seventeen pairs of related subjects (friends, relatives, or partners; 12

male, 22 female; age 23.2 ± 1.9 years; mean duration of relationship 4.8
± 5.3 years) were recruited via announcement in the local students' job
agency. They were of reportedly good health, unmedicated, and
voluntarily participated in the study for the payment of 17 Euros.
Informed consent was obtained from all subjects.
Stimuli
As visual stimulus, a checkerboard pattern with 64 x 48 black and

white squares (angular size ≈ 0.25° x 0.25°) was presented on a 19'' LCD
monitor. The stimulated subject observed the monitor from ≈ 125 cm
distance (visual angle 16° x 12°). Duration of each stimulation period
was one second, consisting of four 250 ms periods with reversed
pattern presentations (see Figure 1).
interstim. int. --- 250 ms --- 250 ms --- 250 ms --- 250 ms --- interstim. int.
Figure 1. Visual stimulation using checkerboard reversals
133
168 stimuli were presented within each of the two runs of the
experiment according to Wackermann et al. (2004). Interstimulus
intervals were 1.5−7.5 seconds; the sequence of intervals was pseudo-
randomized. During interstimulus intervals, the screen was uniformly
grey with the brightness adjusted to the overall brightness of the
checkerboard stimuli.
Physiological Recording
EEG data were recorded from nineteen electrodes according to the

standard 10-20 positions (Jasper, 1958), against linked mastoids as
reference. Impedances were kept below 5 kΩ; Ag-AgCl electrodes were
attached to the scalp with use of Easy Caps and abrasive gel (both by
Falk Minow, München, Germany). Vertical and horizontal oculogram
was recorded for artifact control. Two Synamps amplifiers (NeuroScan,
Inc.) of the same type were used. EEG was band-pass filtered at 0.15 -
70 Hz, A/D-converted, sampled at 1000 Hz and saved to disc using the
software Acquire, Version 4.3.1 (Neurosoft, Inc.).
Procedure
During the experiment, the subjects were seated in two spatially

separate chambers (distance: 8 meters). The experimental rooms were
acoustically and electromagnetically shielded and located on different
sides of the laboratory main room (Figure 2).
Figure 2. Spatial arrangement of the experimental setup
134
Ambach
Each subject was observed via a video-camera by one of the two

experimenters. While one subject of each pair, randomly assigned, was
stimulated visually in room A, the second was instructed to relax with
eyes open in the dimly lit room B. The whole experiment consisted of
two runs, the 'uncovered' (experimental) and the 'covered' (control)
condition; the sequence of both was balanced over the study. Between
both runs, the experimenter responsible for the subject in room A
entered the chamber and either covered or uncovered the monitor. The
break between both runs was kept between three and five minutes. The
non-stimulated subject was not informed about the change of
conditions.
Data analysis
For the analysis, offline EEG data were filtered to 0.15 - 30 Hz and
visually inspected for artifacts. Data from both stimulated (A) and non-
stimulated (B) subjects were segmented into stimulation and
interstimulus epochs of one second duration, according to subject A’s
stimulation procedure.
Segments contaminated
with artifacts were
discarded from further
evaluation; the average
loss rate was 34%. For
subsequent analyses, the
valid segments were
detrended by subtracting
each segment’s third-
order polynomial
approximation from the
data.
In order to compare
subject B's average EEG
power between subject
A's stimulation and
interstimulus epochs, a
distribution-free boot- Figure 3. Flow chart of EEG data processing
strapping procedure was (from Wackermann et al., 2004)
developed and applied
135
by Wackermann et al (2004). Figure 3 illustrates the procedure as a

flow chart (from Wackermann, 2004).
For each non-stimulated subject, valid stimulation (S) and
interstimulus (I) EEG epochs were collected in two pools. All NS
stimulation epochs were averaged; the EEG power Veff was calculated
for the averaged signal as theroot mean square of voltage ui of the n
1 n 2
single data points: Veff = ∑ui . From the NI interstimulus epochs,
n i =1
10000 subsets were drawn randomly, each comprising NS segments.
Within each subset, the EEG power was computed for the averaged
signal. The rank R of the stimulation epochs EEG power was
determined within the cumulative distribution function of the
interstimulus EEG power values for the 10000 subsets. The procedure
was performed separately for each non-stimulated subject and for each
EEG channel. Ranks R, after division by the number of subsets, were
within the interval [0,1] and reflected the probability of receiving a
smaller value than the actual EEG power value out of an arbitrary
subset of segments by the law of chance. Under the null-hypothesis,
these ranks were assumed to be uniformly distributed between 0 and
1].
The further evaluation steps were adopted from the precedent
study (Wackermann et al., 2004). These comprised the transformation
of ranks into z-values, the collapsing of data over subjects, the
collapsing of data over channels using a chi-square statistic, and the
overall assessment of significance for the study. Two different
evaluations were performed, based on different time windows: The
EEG power was calculated for two different time windows: a) for
whole stimulation epochs of one second and b) for a time window of
136ms around the stimulated subject's maximum EEG response (which
had been determined individually for each stimulated subject and for
each channel electrode position).
For the present analysis, the statistical method was reconsidered
in detail, and certain modifications appeared necessary. These are
described in the following paragraph.
Modification 1: Sampling for the bootstrap
In what follows, the previous sampling procedure turns out to

systematically overestimate effects. From a theoretical point of view,
136
Ambach
this is due to the fact that the method compared one sample
('stimulation' epochs) drawn out of the data population with a second
sample ('interstimulus' epochs) drawn out of the data population by
bootstrapping. This violates the premise of exchangeability of epochs
between sample and reference (Edgington, 1987). As a consequence of
this violation, the second sample, which was used as a reference for the
first, does not adequately reflect the possible variability of the first but
underestimates it. Therefore, the rank of the first mean within the
distribution of means of all subsets is not equally distributed within
the interval [0,1] but tends towards the extremes of the interval. This
inevitably leads to an overestimation of significance of any observed
deviations from randomness in either direction. A simplifying model
simulation which uses numeric surrogate data instead of EEG epochs
and which calculates t-tests for a control of the bootstrap, can illustrate
this phenomenon:
A pool of 1,000,000 standard-Gaussian distributed random real
numbers served as data population. Out of this pool, a first random
sample (sample 1) of N1 = 100 numbers was drawn, for which the mean
μ1 was computed. This was to represent the pool of stimulation period
values, but instead of sets of whole EEG segments (for later averaging
and calculation of EEG power) only one single value μ1 for the mean of
the drawn numbers was obtained.
Then, a second random sample (sample 2) with N2 = 300 numbers
was drawn out of the data population, representing the non-
stimulation segments pool (which is larger, about three to one in
relation, than the stimulation segments pool). Out of this second
sample, 10,000 subsets of numbers were drawn, each as large as sample
1, thus comprising N1 = 100 numbers each. For each subset, the mean
was computed, resulting in 10,000 mean values. The cumulative
distribution function of these mean values allowed determining the
rank R of μ1 (mean of sample 1) within all of the 10,000 subset means.
According to the pre-assumptions of the applied evaluation
method, the rank R should become any value between 0 and 1 with the
same probability (even independently from the distribution of values
within the data population). After transformation of each rank R into a
z-value by the inverse Gaussian function, z-values of multiple
repetitions should be distributed standard-Gaussian.
In order to test this, the procedure of drawing samples as
described above and calculating a rank R for μ1 was done repeatedly
137
100,000 times. Fulfillment of the null hypothesis, meaning that sample

1 and sample 2 were drawn out of the same data population, was
guaranteed by the model.
In parallel to the bootstrap-like method, the first (N1 = 100) and
the second random sample (N2 = 300) were compared for mean
difference by using a t-test as a conventional reference method.
Performing a t-test was regarded as permissible, even with different
sizes of sample 1 and sample 2, since the data were distributed
Gaussian.
Every correct method of testing for mean difference should yield
the same value p, which is the probability that the first sample is drawn
out of a data population with a smaller mean value than the second
sample. Given the correctness of the pre-assumptions, the results of the
bootstrap and the t-test should be identical. The results are depicted in
Figure 4. First, the results obtained by both methods (ranks R obtained
from the bootstrap; p-values obtained by the t-test) are compared in a
scatterplot (Figure 4a); second, the distribution of these values in the
interval [0;1] (pre-assumed to be uniform) is plotted (Figure 4b), and
third, the distribution of the z-transformed values (pre-assumed to be
distributed standard-Gaussian) is plotted (Figure 4c). For the
bootstrap, the pre-assumption is clearly violated, while it is obviously
fulfilled for the t-test. Distribution plots for the bootstrap, but not for
the t-test, show an increased standard deviation (which, according to
pre-assumptions, should be 1.0) and a tendency of results towards the
extremes.
To rectify the observed bias, the sampling procedure has been
modified: the rank R of the stimulation epochs EEG power is now
calculated within multiple subsets drawn from a reference pool
comprising all stimulation and interstimulus epochs of the subject.
This modification accomplishes that the reference pool, now
comprising all epochs, adequately represents the possible variability of
EEG measures obtained from the stimulation epochs. In terms of
preassumptions for the bootstrapping procedure, segments are now
exchangeable. The effect of this modification has been tested with the
aforementioned numeric surrogate data. The results are depicted in
Figure 5 (with an arrangement corresponding to Figure 4). The ranks R
obtained from the bootstrap were now found to be scattered within a
small range around the p-value obtained from the t-test. (Additional
138
Ambach
Figure 4. Original method of the bootstrap vs. Figure 5. Modified method of the bootstrap vs.
t-test: comparison of results from the numeric t-test: Comparison of results from the numeric
simulation. (a) Scatterplot of ranks R (from the simulation. (a) Scatterplot of ranks R (from the
bootstrap) versus p-values (from the t-test). (b) bootstrap) versus p-values (from the t-test).
Distribution of the 100,000 obtained ranks R (b) Distribution of the 100,000 obtained ranks R
and the corresponding p-values yielded by the and the corresponding p-values yielded by the
t-test. (c) Distribution of the ranks R (from the t-test. (c) Distribution of the ranks R (from the
bootstrap) versus p-values (from the t-test) bootstrap) versus p-values (from the t-test) after
after z-transformation, compared with the z-transformation, compared with the standard
standard Gaussian distribution. (Note: the two Gaussian distribution
spikes around 4.0 and -4.0 represent cases in
which the mean of sample 1 was outside the
range of all subset means; instead of using
infinite values; they were summarized at the
edge of the plotted range for illustration)
139
analyses revealed that with an increased number of subsets this range

converged to zero.)
Furthermore, in order to rule out a weakness in test power, an
'effect' of a known size (e.g. 0.1 standard deviations) was simulated.
For this purpose, samples 1 and 2 were drawn out of two different data
populations with the means μ1 and μ2. The difference between μ1 and
µ2 was varied in order to simulate different effect sizes. The results
showed that bootstrap and t-test yielded the same p values, which
suggests that the modified sampling procedure does not bias results in
the case of an existing effect.
Modification 2: Correction for degrees of freedom
A second concern refers to the collapsing of results over EEG

channels. The z-values from the nineteen EEG channels are not
independent from each other; the columns of z-values, each comprising
the results from a single EEG channel for all subjects, are considerably
intercorrelated. This implies that the chi-square resulting from these z-
values has less than nineteen degrees of freedom. Accordingly, the
application of a chi-square test according to the preassumption of
nineteen degrees of freedom (which would be appropriate in the case
of independent channel results) will overestimate the significance of
any observed deviation from randomness. A correction for the degrees
of freedom was achieved by using a Monte-Carlo simulation.
In multiple (5000) permutations of all EEG segments of a subject,
segments were assigned randomly to either the 'stimulation' or the
'interstimulus' epoch pool; pool sizes NS and NI were adopted from the
real study data. This assignment was performed once per subject and
kept constant over the nineteen EEG channels (in order to adopt the
degree of interdependence of the channelwise results from the study
data). For each such permutation, data were analyzed as described
above, which, after collapsing over subjects and over channels, resulted
in a chi-square value for each time window (136 ms, 1000 ms) and each
condition (uncovered, covered, and difference). The cumulative
distribution function for the 5000 collected chi-square values was
plotted. From this curve, for any chi-square value, a corresponding p
value can be determined, which indicates the probability of not
exceeding the according chi-square value by randomness.
140
Ambach
For the chi-square values obtained as study results, the

corresponding p values were readable from the cumulative distribution
functions that resulted from the simulation. This allowed an
assessment of significance which was corrected for the degrees of
freedom of the EEG channels.
Results
All analyses have been performed for two different time

windows, 136 ms and 1000 ms; the results are described for both
'uncovered' and 'covered' conditions, as well as for the difference
between conditions. Channelwise results after collapsing z-values over
subjects are summarized in Figure 6. Chi-square values, resulting from
collapsing data over channels, are included for each time window and
condition.
Figure 6. Results after collapsing over subjects: Channelwise z-values for the 136 ms and the 1000
ms time window: 'uncovered' and 'covered' condition, and 'difference between conditions'. Chi-
square results after collapsing results over channels are included.
141
Figure 7. Cumulative chi-square distributions from the Monte-Carlo simulations; A: 136 ms

window, B: 1000 ms window. Vertical, dashed lines denote the chi-square values obtained from the
study for uncovered and covered conditions and the difference between them. Corresponding p
values for any chi-squares are read from these distributions; vertical, solid lines mark the lower and
upper cutoff chi-squares according to the significance level of 0.05 (two-tailed test). The dotted
curve reflects the cumulative chi-square distribution for nineteen degrees of freedom (reflecting the
evaluation without correction for degrees of freedom)
Monte-Carlo simulations for both time windows were performed

in order to estimate the statistical significance of the chi-squares (i.e.
the probability with which pure randomness would produce larger
effects than the study data). The results from the Monte-Carlo
simulations serve as a reference for the chi-squares obtained from the
142
Ambach
study data, as depicted in Figure 7. It can also be read from the figure,
which excessive significance levels would correspond to the obtained
chi-squares without correction for the degrees of freedom (dotted
lines).
Comparison with the previous method
In order to estimate the effect of the methodological

modifications on the study results, the study data were also analyzed
using a) the previous method without modifications ('original
method'), b) the previous method with change of only the sampling for
the bootstrap, but without correction for degrees of freedom ('new
sampling'), and c) the modified method ('new method'). Chi-squares
and significance levels obtained from a), b), and c) are compared in
Table 1. Note that 'new sampling' and 'new method' do not differ with
respect to chi squares but with respect to significance levels.
Table 1: Dependence of study results on the evaluation method: Chi-squares and significance
levels (p values for two-tailed testing, in brackets) obtained from the original method, obtained
after the modification of only the sampling procedure ('new sampling'), and obtained after
both modifications of the method ('new method')
136 ms 1000 ms
Original New New Original New New
method sampling method method sampling method
Uncovered (u) 21.43 17.05 17.05 33.91 23.31 23.31
(.627) (.827) (.898) (.038) (.448) (.511)
Covered (c) 15.45 11.94 11.94 29.00 27.98 27.98
(.623) (.224) (.303) (.132) (.168) (.297)
Difference (u-c) 30.69 23.21 23.21 49.66 40.99 40.99
(.087) (.456) (.492) (.0003) (.0048) (.067)
A further Monte-Carlo simulation was performed to investigate

the proportion of positive study outcomes with the previous method
under fulfillment of the null-hypothesis. The latter was achieved by
multiple random permutations of all 'stimulation' and 'interstimulus'
EEG epochs of a subject. Strikingly, the evaluation using the previous
method without modification falsely led to 20.1% (26.2%) significant
143
results per condition for the 136 ms (1000 ms) time window, when the
significance level α was actually set to 0.05.
Discussion
The present study on correlations in the EEG of spatially

separated subjects aimed at a replication and possible clarification of
the effects described by Wackermann et al. (2004). The research
paradigm was adopted from there.
The EEG power of the non-stimulated subject did not
significantly differ between stimulation and interstimulus periods of
the stimulated subject. This overall result was obtained after two
modifications of the statistical method: A change in the sampling
procedure for the bootstrapping, and a correction of the chi-square test
for the degrees of freedom of the nineteen EEG channels.
It has to be noted that the chi-square for the 1000 ms time
window and 'difference between conditions' just missed the
significance level and therefore might remain an object of discussion.
On the other hand, a correction of the significance levels for multiple
testing has to be considered, because a number of separate evaluations
and significance assessments were performed a) for two different time
windows and b) for the 'uncovered' and the 'covered' condition and for
the difference between conditions. However, even without a correction
for multiple testing, none of the six chi-squares exceeded the
significance level of 0.05.
The overall result is seen as a negative outcome of the attempt to
replicate prior findings of EEG correlations in this specific
experimental paradigm. Of course, it may not be interpreted as a proof
of the absence of the phenomenon under question; firstly and trivially,
the absence of an effect cannot be proven by an unsuccessful
replication attempt; secondly, the outcome refers to the specific
constellation of EEG correlations if one subject is visually stimulated
by checkerboard reversal, which is an abstract pattern.
It is further suggested that the results presented here should not
be discussed in terms of 'telepathy', which was the motivational origin
of some early studies on EEG correlations between separated subjects
(e.g. Grinberg-Zylberbaum et al., 1994). The recent discussion in the
field rather refers to 'non-local correlations between distant brains'
which is a broader concept than the assumption of a directed
144
Ambach
transmission of information by means of telepathy. Regardless of the

viewpoint however, the present study does not contribute to
substantiate the assumption of unexplainable non-local correlations.
Besides the failed replication, the study critically revised the
statistical procedure applied in previous studies. It revealed the
necessity to introduce two important changes.
With the previous method being applied without modification,
results for the 1000 ms (whole-epochs) evaluation suggested a
significant change in the non-stimulated subject's EEG power in the
'uncovered' condition, with a highly significant difference between
'uncovered' and 'covered' conditions. The change consisted of an
increase in EEG power in parietal, predominantly left-hemispheric
regions and a decrease in bilateral temporal regions, which is quite
different from the topography of the effects reported by Wackermann
(2004). However, using the modified analysis, all observed changes in
EEG power were below significance level. Therefore, the results
obtained from the previous method are likely due to inadequate
statistical analyses.
It is expected that the results of those preceding studies, which
used the statistical method discussed here without the two proposed
modifications, and which claimed effects unexplainable by
conventional physiology and physics, will undergo a decrease in effect
size and significance when the revised method is applied on these
data. From here, it cannot be estimated with precision, to what extent
the previous results have been biased by the statistical method, and to
what extent either of the modifications would affect these previous
results. However, both of the methodological annotations must have
led into the same direction, namely to overestimate the effects under
investigation.
References
Charman, R.A. (2006). Has direct brain to brain communication been demonstrated
by electroencephalographic monitoring of paired or group subjects? The Journal
of the Society for Psychical Research 70, 1-24.
Edgington, E.S. (1987). Randomization Tests. New York, NY: Marcel Dekker Inc.
Fenwick, P., Vigus, N., and Sanders, S. (1998). The transferred potential. Unpublished
manuscript, London.
145
Grinberg-Zylberbaum, J., Delaflor, M., Attie, L., and Goswami, A. (1994). The
Einstein-Podolsky-Rosen Paradox in the brain: the transferred potential.
Phys Essays, 7, 422-427.
Jasper, J.J. (1958). The 10/20 international electrode system. Electroencephalogr. Clin.
Neurophysiol., 10, 371-375.
May, E. C., Spottiswoode, S. J. P., and Faith, L. V. (2001). A methodological issue in
the study of correlation between psychophysiological variables. Proceedings
of the 44th Annual Convention of the Parapsychological Association, 166-178.
Sabell, A., Clarke, C., and Fenwick, P. (2001). Inter-subject EEG correlations at a
distance - the transferred potential. Proceedings of the 44th Annual Convention
of the Parapsychological Association, 419-422.
Standish, L. J., Kozak, L., Johnson, L. C., and Richards, T. (2004).
Electroencephalographic evidence of correlated event-related signals
between the brains of spatially and sensory isolated human subjects. The
Journal of Alternative and Complementary Medicine, 10, 307-314.
Radin, D.I. (2003). Thinking outside the box: EEG correlations between isolated
human subjects. Proceedings of the 43th Annual Convention of the
Parapsychological Association, 184-199.
Radin, D. I., (2004). Event-related electroencephalographic correlations between
isolated human subjects. The Journal of Alternative and Complementary
Medicine, 10, 315-323.
Wackermann, J., Seiter, C., Keibel, H., and Walach, H. (2003). Correlations between
brain electrical activities of two spatially separated human subjects.
Neuroscience Letters, 336, 60-64.
Wackermann, J. (2004). Dyadic Correlations between Brain Functional States:
Present Facts and Future Perspectives. Mind and Matter, 2, 105-122.
Wackermann, J., Naranjo Muradás, J.R., and Pütz, P. (2004). Event-Related
Correlations between Brain Electrical Acitivites of Separated Human
Subjects: Preleminary Results of a Replication Study. Proceedings of the 47th
Annual Convention of the Parapsychological Association, 465-468.
146
European Journal of Parapsychology c 2008 European Journal of Parapsychology

Volume 23.2, pages 147–153 ISSN: 0168-7263
Dyadic EEG correlations re-examined:

A commentary on the replication study
by W. Ambach
Jiřı́ Wackermann
Department of Empirical and Analytical Psychophysics
Institute for Frontier Areas of Psychology and Mental Health
Freiburg i. Br., Germany
Some years ago we were reporting results of an electrophysiologi-

cal study (Wackermann et al., 2003) suggesting a relation between brain
states of two separated subjects, assessed by parallel EEG measurements
from both subjects. In an attempt to replicate the finding with an im-
proved experimental setup (Wackermann et al., 2003), we again found
deviations of our measure of interest from the null-hypothesis expecta-
tion, but showing a confusing (if any) pattern difficult to reconcile with
our earlier findings. We were facing a situation indicating either a ‘para-
normal’ nature of the alleged effect, or its possibly artefactual origin. A
follow-up study in our laboratory could not be carried out as planned
due to personnel changes in our department. Therefore, the opportu-
nity to arrange a replication study in co-operation with the research
group for Clinical and Physiological Psychology was welcomed. True
replication studies are rather rare; now there was an excellent opportu-
nity to test whether the same (or at least qualitatively similar) results can
be obtained in an independent study, in another laboratory and with a
different equipment, but following an identical experimental protocol.
The results of the replication study conducted by Dr Ambach are pre-
sented as a separate paper (Ambach, 2008), and we shall just add a few
comments.
———
Address correspondence to: Jiřı́ Wackermann, Institut für Grenzgebiete der Psychologie und Psy-
chohygiene, Wilhelmstr. 3a, D-79098 Freiburg, Germany. E-mail: jw@igpp.de .
147
Wackermann
Firstly, some introductory remarks for a reader who may not be

familiar with the research topic and its history. The problem under
study was introduced by Grinberg-Zylberbaum & Ramos (1987) and
Grinberg-Zylberbaum et al. (1994) who were searching for physiolog-
ical indicators of hypothetical non-sensory, ‘direct communication’ be-
tween two subjects. The term ‘correlation’ used in the title of the com-
mented paper and our earlier communications does not refer to cross-
correlations between native EEG signals. Rather, it refers to correla-
tion between brain states (see Wackermann, 2004), induced by sensory
stimulation at one of the subjects (A), and possibly revealed by a suit-
ably chosen signal parameter in the brain electrical activity of the other,
non-stimulated subject (B). Stimulus-onset aligned data averaging is
a standard technique used in studies of stimulus-related brain electri-
cal responses (also called ‘evoked potentials’, EP). The experimental
paradigm used in our studies was derived from that used by Grinberg-
Zylberbaum et al. (1994), who reported an EP-like activity in the subject
B’s averaged EEG recorded synchronously with subject A’s stimulation.
We could not confirm Grinberg-Zylberbaum’s finding of a ‘transferred
potential’, but we observed subtle power fluctuations of the averaged
EEG signal (Wackermann et al., 2003).1
Specifically, the parameter of interest was the effective voltage ra-
tio, Q ≡ Veff /Vref , where Veff was the effective voltage determined for
the averaged NS EEG data segments recorded during the visual stimula-
tion of the other subject (briefly: stimulation segments), and Vref was a
reference value determined from the same subject’s EEG data recorded
during the inter-stimulation periods. The data-analytic strategy applied
in our studies relied upon an elementary theorem of probability theory
(Feller, 1950): Let X be a continuous random variable with a cumulative
distribution function (CDF) FX ( x ) ≡ P( X < x ); then the transformed
variable U = FX ( X ) is uniformly distributed on the interval [0,1], so the
CDF of U is FU (u ) ≡ P(U < u ) = u; a principle underlying all ‘tests of
significance’.2 Since we had no prior knowledge or a parametric model
1 The exclusive focus on averaged EEG data was occasionally and deservedly criticised. It should
be clear from the above-said that this choice was historically determined: if we were designing a data-
analytic approach from scratch, certainly we would not begin with data averaging. This observation
also reveals limitations of the empiricist approach, building experiments by variations of others’ exper-
iments rather than on prior theoretical reasoning.
2 Where usually the CDF of a test statistics, say, T, under a given hypothesis is derived analytically
and used to transform the observed values of T to U = FT ( T ) or, by convention, to its complement P =
1 − U. It is the uniformity of the CDF of U what allows us to read the P-values directly as ‘probabilities’
of reaching the observed value of statistics T.
148
Commentary on the Replication Study by W. Ambach
of the distribution of Q, empirical CDFs of Q were constructed from re-

peated random drawings of inter-stimulation EEG data sub-sets from a
pool of a total size NI > NS . These CDFs were used to transform the
observed Q values for the stimulation EEG data, i.e., to determine their
ranks R with respect to a ‘reference population’ of non-stimulation Qs.
[The procedure was summarised in (Wackermann et al., 2004), Fig. 1,
reproduced in the commented paper as Fig. 3.] For the purpose of fi-
nal statistics, it was assumed that, under the null hypothesis, the ranks
R = FQ ( Q) are uniformly distributed on [0,1]. Under this assumption,
values Z = Φ−1 ( R) would be normally distributed.
Ambach’s numerical simulations, based on random drawings from
a Gaussian pool (pp. 8*–9*) show that this assumption not necessarily
holds: the distribution of ranks evaluated with respect to empirical CDFs
may be non-uniform (p. 10*, Figs. 4a).3 The real distribution of Rs shows
a relative excess at the lower and upper ends of the [0,1] interval, which
results in a ∪-shaped PDF reminding of arcsin distribution. The devia-
tion from uniformity as such is not surprising; it can be easily predicted
for small samples from the discrete character of the ranks. What is rather
counter-intuitive — at least for a worker used to rely upon asymptotic
statistics — is that with increasing size of both samples the curvature
does not flatten to a uniform distribution; it only becomes smoother
but preserves its shape. This argument is essentially qualitative: it says
that the deviation from uniformity may in principle bias the rank-based
statistics; the outcome of the simulation is presented ‘as is’, and no ra-
tionale is given for the particular choice of simulation parameters.
It is, however, instructive to examine the distributions of empiri-
cally determined ranks with varied sample sizes. We used the same sim-
ulation procedure as Ambach, with only slightly different sum N1 + N2
= 360, instead of 400, and partitioned the total 360 in varied ratios. The
simulations4 (Fig. 1) reveal that the magnitude of deviation from unifor-
mity depends essentially on the samples size ratio N1 /N2 . For ratio 1/3
(Fig. 1, upper row, middle), the results agree with those reported by Am-
bach; the deviation is even more pronounced for ratio 1/2 (upper row,
middle). With decreasing N1 /N2 ratio, however, the CDF comes closer to
the ideal form (see Fig. 1, lower row), so the deviation of Z-transformed
3 Inour view, the co-lateral argument based on parallel t-tests is rather diverting and obscuring the
central message, i.e., the deviation of ranks distribution from uniformity. We do not need a t-test to
know the theoretically correct distribution.
4 The source C code of the simulation program is available on request from the author.
149
Wackermann
1.0 1.0 1.0
N 1 = 120 N 1 = 90 N 1 = 60
N 2 = 240 N 2 = 270 N 2 = 300
0.8 0.8 0.8

Cumulative probability P(R )

0.6 0.6 0.6
0.4 0.4 0.4
0.2 0.2 0.2
0.0 0.0 0.0

0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
Rank R Rank R Rank R
1.0 1.0 1.0
N 1 = 45 N 1 = 36 N 1 = 30
N 2 = 315 N 2 = 324 N 2 = 330
0.8 0.8 0.8


0.6 0.6 0.6
0.4 0.4 0.4
0.2 0.2 0.2
0.0 0.0 0.0

0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
Rank R Rank R Rank R
Figure 1. Cumulative distribution functions of ranks R determined by numerical simulations

according to Ambach’s procedure, with varied sample sizes N1 (‘experimental’ sample) and N2
(‘reference’ sample). Ratios N1 /N2 are (upper row) 1/2, 1/3, 1/5, (lower row) 1/7, 1/9, and
1/11. The dashed diagonal lines indicate the ideal (uniform) CDF. Note that from N1 /N2 ≈ 0.1
on the deviation is almost negligible.
ranks, Z = Φ−1 ( R), would diminish with N1 N2 . It follows, then,

that the rank-based procedure used in our earlier studies is in principle
sound if the samples size ratio is taken into account. The ratio NS /NI (stim-
ulation vs. inter-stimulation data segments) can be roughly determined
by the experimental design.5 Alternatively, the deviation from the uni-
form distribution could be modelled parametrically and corrected for in
subsequent steps of the analysis.
At any rate, Ambach’s study has pointed out that more attention
has to be paid to the construction of data-analytic methods used in
experimental studies of psycho(physio)logical or biophysical ‘anoma-
lies’. It has demonstrated that apparently ‘intuitive’ statistical proper-
ties must not be assumed without careful verification, especially where
relatively small samples (N ≈ 101 — 102 ) are concerned: an important
aspect that should be appreciated. However, the insights derived from
5 We
say ‘roughly’ because the exact value will depend on the data loss rate due to artifact elimina-
tion, and is thus not perfectly under experimenter’s control.
150
the simulations should not be applied too hastily. In the commented

paper, the bare fact that P-values obtained from empirical distribution
functions by our original method may be biased was turned into an argu-
ment for the necessity of merging the two data subsets (stimulation and
inter-stimulation EEG segments), and for using the combined dataset as
a reference pool (p. 11*) — a point on which we disagree. It is difficult
to resist an impression that the eventually successful ‘deconstruction’ of
the effect under study serves here as a proof of statistical reconsidera-
tions and modifications based thereupon. We would prefer a systematic
study of the bias, focussing particularly on (i) a quantification of the
bias as to its dependence on the dataset size, and (ii) a more convincing
demonstration that the model used in demonstration of the bias applies
to analyses of real EEG data.
For a reader who may find our assessment slightly opinionated we
wish to add that we do not argue in favour of the ‘reality’ of the effect
under question. In fact, from our operationalist stance the phrase ‘re-
ally existing effect’ has no other meaning that ‘consistently reproducible
phenomenon’ (given that the necessary conditions for the phenomenon
are known). Reviewing results of our own replication study (Wacker-
mann et al., 2004), and taking into account the results reported by Am-
bach even without post hoc corrections, it is highly doubtful that there is
anything such as a ‘real’ effect. This negative evidence arises from the
remarkable lack of consistency, in terms of direction and spatial distri-
bution of the effect measures, and is thus based more on a visceral feel-
ing for ‘what is like real physiology’ than on formal statistical inference.
———
Our final comment is not related directly to the commented study
but, generally, to studies of correlations between physiological activi-
ties of separated subjects, or any biophysical ‘anomalies’ of similar sort.
During the last few decades, the search for so-called ‘physiological signs
of psi’ (Beloff, 1974) went through several waves of enthusiasm, of-
ten elicited by novel experimental approaches (‘direct communication’,
‘direct mental interaction between living systems’, ‘presentiment’, etc),
and sometimes claimed to promise a fundamental revision of our view
of nature (e.g., quantum-like ‘entanglement’ between human brains, ‘re-
versal of time arrow’, etc).
As to our knowledge, none of those high hopes has ever been full-
filled, and none of those approaches developed into a really working ex-
151
Wackermann
perimental paradigm — that is, one yielding reproducible results across

laboratories, results that would visibly stand out of the bush of error
bars. There are no signs of real progress. May each individual researcher
draw her/his conclusions from this observation. We take the lesson se-
riously and turn to more productive research topics, not to spend our
lives in a heroically ‘relentless’ but ultimately unproductive search.
Acknowledgements
Thanks to Dr Ambach for intense discussions of the topical issues,
and to two reviewers for helpful remarks on this commentary.
Abbreviations and mathematical symbols

CDF cumulative distribution function
EEG electroencephalogram
EP evoked potential
PDF probability density function
FX CDF of random variable X
Φ CDF of normally distributed random variable
Φ−1 inverse function to Φ
N sample size
P probability of an observed event
P probability associated with a test statistics
Q effective voltage ratio
R rank value
U uniformly distributed random variable
Veff effective voltage (stimulation)
Vref reference effective voltage (inter-stimulation)
X random variable, general
Z random variable, normally distributed
References
Ambach, W. (2008). Correlations between the EEGs of two spatially separated subjects
— A replication study. European Journal of Parapsychology, this issue.
Beloff, J. (1974). ESP: The search for a physiological index. Journal of the Society for
Psychical Research, 47, 403–420.
Feller, W. (1950). An Introduction to Probability Theory and Its Applications, Vol. I. New
York: Wiley.
Grinberg-Zylberbaum, J. & Ramos, J. (1987). Patterns of interhemispheric correlation
during human communication. International Journal of Neuroscience, 36, 41–53.
152
Grinberg-Zylberbaum, J., Delaflor, M., Attie, L., & Goswami, A. (1994). The Einstein-
Podolsky-Rosen paradox in the brain: The transferred potential. Physics Essays 7,
422–427.
Wackermann, J., Seiter, C., Keibel, H., & Walach, H. (2003). Correlations between brain
electrical activities of two spatially separated human subjects. Neuroscience Letters,
336, 60–64.
Wackermann, J. (2004). Dyadic correlations between brain functional states: Present
facts and future perspectives. Mind and Matter, 2(1), 105–122.
Wackermann, J., Naranjo Muradás, J. R., & Pütz, P. (2004). Event-related correlations
between brain electrical activities of separated human subjects: Preliminary re-
sults of a replication study. In S. Schmidt (Ed.), Proceedings of the 47th Annual
Convention, Vienna: Parapsychological Association, pp. 465–468.
153
Lunar-magnetotail Encounters as Modulators of

Mind-Matter Interaction Effects
Eckhard Etzold
Society for Anomalistics, Postfach 1202, D-69200, Sandhausen, Deutschland
Abstract
Mind-matter interaction (MMI) effects are controversial partially

due to a lack of a theory that persuasively explains such effects, and
also due to the difficulty of replicating empirical effects in
controlled experiments. One potential explanation for this
empirical capriciousness might be the presence of physical factors
which modulate MMI performance. One previously suggested
variable has been the Earth's geomagnetic field; another is the lunar
phase. The hypothesis that the Moon's interaction with the Earth's
magnetosphere modulates MMI performance was tested in data
collected in a long-term, online MMI experiment. The analysis
showed a clear influence of the Earth's magnetotail in the MMI
results, confirming the hypothesis. This suggests that phenomena
relying on purported MMI may be more efficacious during quiet
geomagnetic periods of the full moon, when the Moon is passing
deep through the inner plasma sheet of the magnetotail
Introduction
Mind-matter interaction (MMI) is the hypothesized ability of the

mind to influence matter or energy without the use of any currently
known physical means. Spiritual healing, psychokinesis, distant
healing, and unusual human-machine interactions might be based on
MMI effects (Heath, 2003). Although hundreds of laboratory studies
seem to provide evidence of MMI, the existence of such effects remains
controversial because there are no convincing theoretical reasons to
explain how such effects may exist (Bösch, Steinkamp, Boller, 2006;
Radin, Nelson, Dobyns, Houtkooper, 2006). Barring theory, the central
Correspondence details: Eckhard Etzold, Grosse Grubestr. 2A, D-38122 Braunschweig, Deutschland.
Email: eckhard.etzold@gmx.de
154
Lunar-magnetotail Encounters as Modulators of MMI Effects
issue in this line of research has been the question of reliability. For
example, a large-scale, multi-laboratory replication effort failed to
repeat previously successful MMI experiments reported by Princeton
University’s Engineering Anomalies Research Laboratory (Jahn et al.,
2000). While critics might point to the failure to replicate as evidence
that MMI does not exist, another interpretation is that we do not fully
understand the conditions under which MMI is best demonstrated.
This prompted a search for environmental parameters which modulate
the MMI performance via enhancing or reducing human powers of
concentration.
Geomagnetic effects on human behaviour
The influence of subtle electromagnetic effects on our human

nervous system has been a topic of scientific research for decades.
Evidence of an effect of ELF electromagnetic fields on human pineal
gland function was reported (Wilson et al., 1990) and rejected (Gobba,
Bravo, Scaringi & Roccatto, 2006). Persinger found correlations of
geomagnetic activity with enhanced anxiety, sleep disturbances,
altered moods, and greater incidences of psychiatric admissions
(Persinger, 1987). Researchers had reported that poltergeist episodes
frequently begin on the day of a sudden and intense increase in global
geomagnetic activity (Gearhart & Persinger, 1986). Sturrock (2004)
analysed data of UFO events and found significant correlations with
local sidereal time. Spottiswoode (1990) reported the existence of a
negative correlation between scores in free response anomalous
cognition experiments and geomagnetic fluctuations. Sudden infant
deaths seem to be associated with continuous micropulsations in times
when global geomagnetic activity is very low (O'Connor & Persinger,
1999). Increased solar and geomagnetic activity seems to be associated
with increased arterial blood pressure (Ghione, Mezzasalma, Del
Seppia & Papi, 1998). Distinct effects of the Moon on geomagnetic
activity were also reported (Bigg, 1963; Bell and Defouw, 1964; Knott,
1975).
Despite the large volume of reports, in total we do not really
understand the role of these electromagnetic and geomagnetic
interactions on human behaviour. Neither do we understand the
underlying mechanism of MMI effects. As a result, attempts to explore
MMI effects by using patients as targets might raise serious ethical
155
Etzold
problems, since we don’t know when MMI effects might be healing or

harming others. All that have today is nothing more than some small
pieces of a jigsaw. The use of MMI experiments with a random number
generator as a target for the evaluation of the effectiveness of global
geomagnetic activity on human behaviour does minimize the risk of
adverse effects on the health of subjects and is thus ethically safe. It is a
great way to learn more about this still unexplored realm of human
ability and behaviour.
Lunar phase effects in MMI data
In an earlier study, Radin (1997) claimed evidence of MMI effects

in casino pay-out rates which depended on the lunar phase. The peak
effect was found within one day of the full moon. Based on many
previous studies examining correlations between the strength of the
Earth’s geomagnetic field (GMF) flux and purported psychic
experiences, it was predicted that the casino pay-out rates would be
negatively correlated with GMF. Radin found a higher than average
payout rate on full moon days of quiet geomagnetic activity, and a
lower than average payout rate on full moon days with high
geomagnetic activity. In both cases yet, the correlation would be
negative. Thus, we may assume that the effects might cancel each other
out.
These results might indicate that the human power of
concentration is affected by a geomagnetic parameter. Recently,
Sturrock and Spottiswoode (2007) reported a significant lunar-periodic
effect in free response anomalous cognition experiment data, which
might also confirm this relationship of a lunar-geomagnetic parameter
and the human power of concentration.
Attempts to independently check Radin’s claims using data from a
large-scale, online MMI experiment revealed evidence for a complex
solar-periodic full moon effect (Etzold, 2000; Etzold, 2002).
The Fourmilab Retropsychokinesis Project
Etzold (2005) analysed data of the Fourmilab RetroPsychoKinesis

Project, an online experiment which “explores the purported
anomalous effect known as retropsychokinesis”1, also known as
1
http://www.fourmilab.ch/rpkp/proposal.html
156
“retroPK“. RetroPK is a hypothesized ability to retroactively influence

random data via MMI. Another hypothesis, the Decision
Augmentation Theory (DAT), attempt to explain retropsychokinesis
effects with an anomalous information transfer by decision
augmentation. This means that people are somehow able to foresee the
future and intuitively feel which goal of the experiment might be more
promising (May, Utts & Spottiswoode, 1995). If the DAT hypothesis is
true, one would expect in case of strong MMI effects in the
experimental data a reverse effect in control data taken from the same
database. Since DAT can be reduced to intuitive data selection, this
might be in fact a kind of data splitting. The lack of any reverse effects
in control data might indicate that the hypothesis of retroactive
influence is more qualified to explain the effects than intuitive data
selection.
Random data for the Fourmilab test are derived from the HotBits
hardware random number generator (RNG) based on radioactive
decay2. Date and time stamps for each experimental result,
downloaded from that website, were converted to a lunar phase in
degrees. 105 Subsets or bins were created for each lunar cycle from
January 1997 to October 2005 (Etzold, 2005). These MMI data were
correlated with F10.7 solar radio flux, sunspot numbers, solar wind
speed and GMF ‘ap’-index data. Significant results were found for
MMI data correlated with F10.7 solar radio flux and sunspot numbers.
The correlation of the GMF ‘ap’-index data with MMI data was barely
significant on the p = .05-level. This barely significant result might
depend on remaining diurnal variations in the GMF ‘ap’-index data.
Therefore we tried another analysis with reduced diurnal variations in
the GMF ‘ap’-index data, and we expected an increasing significance
for the correlations of the GMF ‘ap’-index data with MMI data.
The Moon-magnetotail interaction hypothesis
A hypothesis was suggested by earlier analyses that the Moon’s

interaction with Earth’s magnetosphere during the Moon’s passage
through the magnetotail in full-Moon times might modulate MMI
performance (Etzold, 2005). We wanted to test this hypothesis and see
whether high and low MMI effects in the full-Moon interval could be
explained by changing geomagnetic activity.
2
http://www.fourmilab.ch/rpkp/experiments/contents.html
157
Etzold
While the effective parameter responsible for modulating MMI

performance remains unknown, analyses seemed to indicate that the
origin of this parameter may be located in the lunar phase. We
assumed more precisely that the actual source of this parameter may
be based on lunar-magnetotail interactions.
As illustrated in Figure 1, the Earth’s magnetosphere is distorted
by the solar wind into a teardrop shape, which extends far into deep
space (Tsyganenko, 1995). The magnetotail of this drop stretches in a
direction opposite to the Sun, and reaches beyond the lunar orbit. The
solar wind follows a path that goes around the Earth in a sort of cover
or sheath. This is the magnetosheath plasma region. Between the
magnetosphere and the magnetosheath is the magnetopause. The
magnetotail is divided into the outer lobes and the inner plasma sheet.
During the full moon interval, the Moon passes the
magnetosheath plasma and the lobes of the magnetotail for
approximately four days (Schubert, Sonett, Smith, Colburn &
Schwartz, 1975).
An indicator for the extent that the Moon spends in the
magnetotail is the time the Moon spends in the inner plasma sheath
(figure 2). This time varies from 0-15 hours per month at a minimum
rising to 60-75 hours per month at the maximum with the 18-year Saros
period (Hapgood, 2007) and with the varying solar activity
(Tsyganenko and Sitnov, 2007) of the 11-year solar cycle. The lunar
orbit is tilted in relation to the ecliptic, and in this way the Moon can
pass above, below, or directly through the plasmasheet, depending on
the Moon’s position in the Saros period.3 The last period of maximum
length of time the Moon spent in the plasmasheet was from 1993 to
2001, and the next period of maximum length is expected around 2011
to 2019. These are very promising time intervals for studying
fullmoon-MMI effects. From 2003 to 2010 there is a period of minimum
length, leading to the expectation of low fullmoon-MMI effects. But
even when the Moon is expected to pass through the magnetotail, a
sudden collapse of the entire nightside field with the magnetotail can
happen when a strong geomagnetic storm is triggered by high solar
activity.
3
An animated .gif-file created by M. Hapgood, showing the lunar crossing of the magnetotail
around the full moon in December, 2007, is available at
http://uk.geocities.com/mike.hapgood@btinternet.com/moon_movie_6.gif.
158
Figure 1. Earth’s magnetosphere. Picture was provided by NASA.
In view of the hypothesis that an unknown parameter is

modulating MMI performance which results from the Moon’s
interaction with the magnetotail, one might assume that the correlation
of MMI data with GMF ‘ap’-index data might yield more significant
results. But this was not the case in the Etzold study (2005). Higher
significance was found for perihelion parameters like F10.7 solar radio
flux and sunspot numbers. The current paper will show that this lack
of significance is due to the properties of the GMF ‘ap’-index data.
Therefore, the data in the present analysis were retrieved from the
same retroPK database for the same evaluation period as mentioned
before. These are all MMI data from the time period of January 11,
1997, 17:33 UTC through October 8, 2005, 12:45 UTC, all of which fulfil
the specifications of the Fourmilab RetroPsychokinesis Project. Newer
data were ignored and retained for future replication attempts.
159
Etzold
Figure 2. Hours, the Moon spends in the inner plasmasheet. Detail of a plot by Mike
Hapgood (2007), and figure was treated by the author. Bright areas mark time
intervals of deep lunar encounters with the inner magnetotail, and with a high
fullmoon-MMI effect expectation.
Methods
We wanted to see what the solar-periodic full moon effect looks like
because the lunar-GMF-MMI hypothesis stated that the Moon’s
interaction with Earth’s magnetosphere during the Moon’s passage
through the magnetotail at full moon times might modulate MMI
performance. We therefore scanned the data with a sliding window
function over the whole lunar phase. We used the GMF ‘ap’-index data
which provide the average global geomagnetic field activity over a
three hour period. The GMF ‘ap’-index data were obtained in a fixed
time interval of every three hours. This results in a constant number of
GMF ‘ap’-index data for every day of the successive lunar cycles.
160
Three different data bases are available on the Fourmilab server:

MMI ‘for the record’ data, MMI ‘practice’ run data and control run
data. Here, we are analyzing the MMI ‘for the record’ data and also
‘practice’ run data and control run data for a comparison.
The properties of the data
The MMI ‘for the record’ experiment data were only recorded
when somebody was conducting an experiment. Thus, the amount of
experimental data varied largely for the same interval of successive
lunar cycles with a minimum of 152 and a maximum of 5,555 ‘for the
record’ experiments per lunar cycle. In total, 199,632 ‘for the record’
experiments and 202,958 practice run results for the experiment time
interval of January 11, 1997 to October 8, 2005 are evaluated. Since
December 26, 1997 control experiments have also been running
automatically, exactly one run every hour, totalling 59,860 as of
October 8, 2005. The first and second generation HotBits configuration
used a Windows 95 machine which “crashed every couple of months
for no discernible reason“4. As a consequence, the Hotbits server was
upgraded in September 2006 with a far more reliable Linux-based
server (Fedora Core 5).
Further complicating matters is the fact that the control runs do
not really match the MMI data. While there are 199,632 MMI
experimental data in the evaluation period, only 59,860 control run
data are at hand which are a little bit more then a third of the MMI
data volume. In addition, practice run results were usually not
reported. But due to the important issue of control data, we have to
consider every source of information and as a result we looked at the
available practice run results.
GMF ‘ap’-index data and diurnal effects
The GMF activity varies in diurnal rhythm and with the location
on Earth. Besides day and night changes, another reason for diurnal
variations is the fact that Earth’s geomagnetic poles do not conform
with the Earth’s geographic poles. As a result, Earth’s magnetotail
shifts in diurnal rhythms. However, the GMF ‘ap’-index data represent
a global average of GMF activity as an integer value with minimized
4
http://www.fourmilab.ch/hotbits/hardware3.html
161
Etzold
diurnal rhythm. For example, the exact full moon time might be in one
lunar cycle at noon (e.g., 12:00 UTC) with a lower GMF level and in the
evening (e.g., 18:00 UTC) of another lunar cycle when the GMF level is
higher. Both time intervals might yield the same GMF ‘ap’-index data
value, but the effect of the lunar/GMF interaction might be completely
different. Etzold (2005) computed the correlations of MMI data with
GMF ‘ap’-index data for every lunar cycle separately. For our
evaluation here this means that we have to combine the data of
multiple lunar phases in order to minimize diurnal effects and make
the data compatible. This means that effects, by accumulation, become
evident which are not present if we evaluate data of single lunar
phases with much smaller volumes of data. Eight GMF ‘ap’-index data
values were obtained every day, and reduced diurnal effects in the
MMI data can be expected if no fewer than four lunar cycles were
included in each subset. This is the case if we divide the entire database
into 16 subsets, each with 12,500 MMI ‘for the record’ experimental
data, and the last with 12,132 experimental data. Date and time stamps
of these data were converted into lunar phase in degrees and sorted
according to ascending lunar phase for each subset.
The control run, practice run, and GMF ‘ap’-index data were also
divided into 16 subsets but with a different number of experiments per
subset with respect to the time intervals of the MMI data to enable
comparability of the results. Date and time stamps of these data were
also converted into lunar phase in degrees and sorted according to
ascending lunar phase for each subset.
Configuring the full moon interval & tests with different sliding window widths
The time interval which the Moon spends in the magnetosphere is

estimated at four days. Radin and Rebman (1998) mentioned an
interval of one day within the full moon period in which the full moon
effect was found in the casino data. Including the day before and the
day after the full moon day, this constitutes an interval of three days
centered around the time of the full moon. On one day the Moon
moves 12.19° along its orbit on average. Thus, Radin’s full moon
interval is equivalent to 161.7° to 198.3° lunar phase. Etzold (2000)
mentioned a lunar phase interval of 166.5° to 192.4° of approximately
two days. Since we compute time intervals in lunar phase in degrees,
we have 25° lunar phase as a minimum for a two-day interval and 50°
162
lunar phase as a maximum for a four day interval. For finding the
maximum magnitude of an assumed full moon effect in the MMI data,
we check all data in steps of 5°, beginning with 25° and ending with
50°. This different window size test should ensure that the effect is not
sensitive to the percentage of the lunar cycle that is analysed.
The data of each MMI subset and the corresponding GMF ‘ap’-
index data subset were sorted according to the the increasing lunar
phase with overlapping data for the new moon interval. A computing
function with a window width of 25° up to 50°, starting with 0° and up
to 360°, overlapping in the start and end-interval, scanned
simultaneously the first MMI subset and the corresponding GMF ‘ap’-
index data subset synchronously from the beginning to the end and
calculated the correlation value for every step (0.5°). The sliding
window function computed the MMI average effect size and the
average GMF ‘ap’-index value for every step. For these values,
Pearson’s r was computed. Since we have 16 subsets, we get 16 Pearson
correlation values for each step of the sliding window function. The
following checklist displays the basic steps of the analysis:
1. Download the MMI data from the Fourmilab server1.

2. Convert UNIX time into UTC date and time, and hexadecimal
experiment data into a bit score.
3. Divide the dataset into ‘for the record’ and practice data.
4. Check databases for data which do not conform with the
Fourmilab Retropsychokinesis Project specifications.
5. Create 16 subsets with 12,500 MMI experimental data lines each.
6. Convert date and time of each experiment into a lunar phase
degree with 0° for new moon, 90° for the first quarter, 180° for full
moon and 270° for the last quarter.
7. Sort data according to increasing lunar phase for each subset,
starting with 0° and ending with 360°
8. Copy data with lunar phase < 60° and add 360 for every subset.
Insert these data at the end of each subset for overlapping scans at new
moon position.
9. Perform the same procedure for control run and practice run data.
10. Download GMF ‘ap’-index data5.
5
The GMF 'ap'-index data were retrieved from the World Data Center for Geomagnetism, Kyoto.
http://swdcdb.kugi.kyoto-u.ac.jp/
163
Etzold
11. Eight GMF ‘ap’-index data were obtained per day. Complete hour
and minute information of the middle of each three-hour ‘ap’-index
interval.
12. Take date and time of the beginning and the end of each MMI
data and divide the GMF ‘ap’-index data into corresponding subsets.
13. Do steps 6 to 8 with GMF ‘ap’-index data.
14. Create a sliding window function which computes Rosenthal’s
effect size for the MMI, practice and control run data and the average
values for the corresponding GMF ‘ap’-index data.
15. Adjust the window width of the sliding window function (25°,
30°, 35°, 40°, 45°, and 50°).
16. Move the sliding window in steps of 0.5° over the entire lunar
phase synchronously over all subsets.
17. Correlate the effect size values with the average GMF ‘ap’-index
data for every sliding window step.
18. Plot the Pearson correlation values.
Tests with split data
A split data test checks that the effect found is not an artifact of the
precise sequence of recorded data. Therefore, the relevant data set was
divided randomly into two samples. If there is a non-random effect in
the data it will be present in both data sets. Two empty data samples
were generated for the data split. A pseudo random function with
current time as a random seed returned a number between 0 and 1 for
every data split. If the pseudo random number was > 0.5, data were
saved in sample #1, otherwise in sample #2. In this way, we obtained
two samples which are independent of each other. The window width
of the sliding window function was set to 36.6°, which is similar to the
range of the full moon interval defined by Radin. Both samples were
correlated with the GMF ‘ap’-index data set. For both data sets we ran
the sliding window function across each data set in steps of 0.5°,
resulting in two rows of 16 average GMF ‘ap’-index data values and 16
corresponding effect size per trial values for every 0.5° step. For this
pair of 16 values each, Pearson’s r with N = 16 was computed, so we
got 720 Pearson correlation values for one sliding window function test
across the whole lunar cycle.
164
Results
The overall results for the total of 199,632 ‘for the record’
experiments yielded a per trial effect size of .49997 (where 0.5 is
expected by chance in this test) which is equivalent to a nonsignificant
z score of -0.82. The control runs yielded a per trial effect size of .50038
which is equivalent to a nonsignificant z score of 0.82, and the practice
runs yielded a per trial effect size of .50001 which is equivalent to a
nonsignificant z score of 0.32. But we have already seen that Radin
found a full moon effect in the casino data with a higher payout rate
within the days of the full moon, and we can assume that the total
payout rate of the casino might not be significantly above the mean
chance expectation. Otherwise, casino holders might worry about their
future. If Radin’s full moon effect is based on MMI, we can expect a
similar effect in the Fourmilab Retropsychokinesis data.
Control run and practice data correlation results
A test with randomly split data was executed for the 36.6° full
moon interval (161.7° to 198.3°). The control and practice run data
yielded non-significant correlation values. For the first control run
sample r(14) was -.15, and for the second sample r(14) was -.09. For the
first practice data sample r(14) was .01, and for the second sample r(14)
was .13. Thus, we might assume that there is no periodical bias in the
data which could generate the full moon effect found in the MMI data.
Figure 3 shows the results for the control run data.
The control runs showed a random walk of the curves. No full
moon effect and no other coherent effects are visible in this figure. This
seems to indicate that the HotBits hardware random number generator
is working well for our purposes here.
MMI ‘for the record’ data correlation results
The test with randomly split data yielded the highest correlation
value for MMI data. The first sample data yielded r(14) = -.62 with p
= .005, one-tailed, and the second sample data yielded r(14) = -.54 with
p = .015, one-tailed. Both datasets showed strong negative correlations
which might indicate that there is a non-random trend in the data.
Figure 4 shows the results for the MMI ‘for the record’ data. By visual
165
Etzold
inspection, both curves showed a similar peak in the full moon interval
which reached significance. This might indicate that there is a non-
random trend in both datasets which is lunar phase dependent.
Figure 3. Smoothed epoch analysis for the split control run data vs. GMF ‘ap’-index data. Data set A:
grey line; data set B: black line.
Figure 4. Smoothed epoch analysis for the split MMI ‘for the record’ experiment data vs. GMF ‘ap’-
index data. Data set A: grey line; data set B: black line.
166
Correlation results of non-split data & with different scanning window widths
We wished to plot the shape of the full moon effect and therefore
the entire non-split MMI database was scanned with different sliding
window widths of 25°, 30°, 35°, 40°, 45°, and 50°, starting at 0° through
360°, moving in steps of 0.5° for each correlation value. Figure 5 shows
the results:
We found a sharply-defined full moon correlation effect which is
visible for all scanning window widths. The peak effect was found at
3.5° before full moon with r(14) = .804 and 40° window width. The
shape of the full moon peak resembled an imprint of Earth’s
magnetotail in the correlation data. A comparison with other Moon
phase intervals showed that this strong negative correlation only exists
for the full moon interval.
Figure 5. Epoch analysis for the retro-PK ‘for the record’ experiment data vs. GMF ‘ap’-index data
with different scanning window widths.
Table 1 lists the correlation values of the full moon MMI data for
different window widths, centered at full moon and for comparison,
the correlation values for the remaining non-full moon data.
167
Etzold
Table 1. Pearson Correlation values for 180° lunar phase (full moon)
and the remaining non-full moon data with n = 16.
Window Pearson correlation of Pearson correlation of

width full moon data non-full moon data
25° -.46* -.030
30° -.59** -.026
35° -.69** -.008
40° -.76*** .008
45° -.69** .008
50° -.62** .066
* p < .05; ** p < .01; *** p < .001 (all probabilities one-tailed)
All MMI data of the full moon intervals yielded significant results.
For Radin’s exact full moon interval of 161.7° to 198.3° lunar phase, we
got r(14) = -.74 with p = .0005, one-tailed due to Radin’s prediction of a
negative correlation. The subset with the lowest average GMF ‘ap’-
index data value of 8.41 yielded a MMI per trial effect size of .5008 with
a standard error of .00043, and the subset with the highest average
GMF ‘ap’-index data value of 29.1 yielded a MMI per trial effect size
of .4993 with a standard error of .0004.
With a window width of 25° we found the initial point of the full
moon correlation effect at 153° lunar phase and the end point at 198°
lunar phase. With a window width of 50°, the initial point is at the
position of 155° lunar phase and the end of the full moon effect at 208°
lunar phase. This is the range in which we expect the Moon’s transit
through the magnetotail of the Earth. The Moon might pass the tail
magnetopause between 198° and 208° (mean = 203°) and between 153°
and 155° (mean = 154°) lunar phase, based on the observations here.
The resulting average width of the Earth’s magnetotail in the lunar
orbit might be 49° lunar phase or a time interval of 4 days.
The center of the magnetotail in the lunar orbit might not be the exact full moon
time, but 6 hours and 53 minutes before the exact full moon time. This might be in
accordance with the expected shape of Earth’s magnetotail due to the fact that the
magnetotail is blown over by the interplanetary magnetic field.
Discussion
The tests in the current study seem to confirm previous observations

of MMI-lunar relationship dependent on geomagnetic activity. If we
168
assume that medical treatment such as distant healing or anomalous

healing depend on MMI, then we may expect that these medical
treatments might also be modulated by lunar-GMF effects, especially at
full moon times of low GMF activity. But this is contrary to other
findings. For example, Palmer, Baumann and Simmonds (2006) studied
the magnetic influence on MMI effects whilst influencing the
hemolysis of red blood cells, and wrote with respect to their study:
“The most sobering implication of our data ... derives from the
evidence of hemolysis acceleration. Translated into healing terms, this
means that healers could unintentionally ‘mis-direct’ their PK to make
an illness worse rather than better. ... If the finding with the Ap index
continues to hold up, it may suggest that healing should be performed
on a day following that when the global GMF is relatively high.“ The
findings in the current study suggest that this statement is too general.
The actual situation seems to be more complex. Healers might perform
better on days of low GMF activity in full moon times with deep
plasmasheet encounters. But if they try to heal at full moon times of
extreme high GMF activity, they could make an illness worse rather
than better because MMI might be active, but working against their
intention.
An important observation in this analysis might be the difference
between MMI ‘for the record’ data and the practice data for the full
moon interval. While a remarkable full moon effect was found in the
MMI ‘for the record’ data, it is completely absent in the practice run
data. Due to the fact that practice and MMI ‘for the record’ data were
taken from the same random source, one might assume that the
difference might be due merely to psychological factors. The missing
full moon effect in the practice data might indicate that a non-serious
attempt to “influence” the output of the RNG might fail to produce a
full moon effect. It might be dependent on the strength or gravity of
intent. Practice alone does not yield a result. Translated into healing
terms, this means that praying, wishing and healing without a serious
intent might not yield a result.
Another point is the accurate shape of the full-Moon effect. We don’t
know what happens when the Moon is passing through the
magnetotail but we may assume that a physical parameter is
influenced which has a strong effect on the human mind to influence
matter without the use of any currently known type of physical
mechanisms. The current findings suggest that MMI might not be an
169
Etzold
elusive and weak effect but rather a stronger and more reliable effect if
we knew the crucial parameters and could apply them systematically
in MMI research.
We also do not understand the exact structure of Earth's farther
magnetosphere. What we have are reconstructions and theoretical
models of the geomagnetic field as observed by spacecraft data
(Tsyganenko, 1995). These spacecraft are located at a distance of 3 to 40
Earth radii maximum (Tsyganenko and Sitnov, 2007). If the sources of
this MMI effect observed here are based upon interactions of the Moon
with the Earth’s magnetotail, then we have a well-formed imprint of
the farther magnetosphere which, beyond the use of expensive
spacecraft, give us empirical clarification about the location of the
magnetopause and the shape of the farther magnetotail at a distance of
about 60 Earth radii. It seems to be unique in parapsychological
research that an unusual MMI effect might be able to operate as a low
cost probe for geomagnetic tail research.
While the casino data in the Radin study were generated at the same
time and in the same location where they had been “treated”, the MMI
data had already been generated at an unknown time before being
used in the Fourmilab RetroPsychokinesis Project. They were delayed
for hours, days, or perhaps weeks, and they were generated far away
from the participants’ location. In both cases, we noticed an unusual
effect, phase locked with the synodic lunar phase. If the retroactive
psychokinesis effect really exists (Schmidt, 1976) ‒ and the lack of any
reverse effect in control and practice data support this assumption ‒
we would not expect an effect in the parts of the lunar phase where the
random data had been generated beforehand, but in this part of the
lunar phase where the test persons observe the data in the MMI
experiments. This was exactly what we had found. This observation
might provide evidence of an ability of the human consciousness to
influence targets distributed in space and time without being limited
by the constraints of the local bodily presence (Schmidt, 1981).
References
Bell B., and Defouw R.J. (1964). Concerning a lunar modulation of geomagnetic
activity. Journal of Geophysical Research, 69, 3169-3174.
Bigg E.K. (1963). The influence of the Moon on Geomagnetic disturbances. Journal
of Geophysical Research, 68, 1409-1413.
170
Bösch H., Steinkamp F., and Boller E. (2006). Examining psychokinesis: The
interaction of human intention with random number generators: A meta-
analysis. Psychological Bulletin, 132, 497-523.
Etzold E. (2000). Lunarperiodische und solarperiodische Einflüsse in
Psychokineseversuchen [Lunar-periodic and solar-periodic effects in
psychokinesis experiments]. Grenzgebiete der Wissenschaft, 49, 149-174.
Etzold E. (2002). Sind die Daten der Fourmilab-Experimente mit der Mondphase
korreliert? Ein Replikationsversuch [Are the data of the Fourmilab experiments
correlated with Moon phase? An attempted replication]. Zeitschrift für
Anomalistik, 2, 76-90.
Etzold E. (2005). Solar-periodic Full Moon Effect in the Fourmilab
RetroPsychokinesis Project Experiment Data: An Exploratory Study. Journal of
Parapsychology, 69, 233-261.
Gearhart, L., and Persinger M.A. (1986). Geophysical variables and behaviour:
XXXIII. Onsets of historical and contemporary poltergeist episodes occurred
with sudden increases in geomagnetic activity. Perceptual and Motor Skills, 62,
463-466.
Ghione S., Mezzasalma L., Del Seppia C., and Papi F. (1998). Do geomagnetic
disturbances of solar origin affect arterial blood pressure? Journal of Human
Hypertension, 12, 749–754.
Gobba F., Bravo G., Scaringi M., and Roccatto L. (2006). No association between
occupational exposure to ELF magnetic field and urinary 6-sulfatoximelatonin
in workers. Bioelectromagnetics, 27, 667-73.
Hapgood M. (2007). Modelling long-term trends in lunar exposure to the Earth's
plasmasheet. Annales Geophysicae, 25, 2037-2044.
Heath P.R. (2003). The PK Zone. A Cross-Cultural Review of Psychokinesis (PK).
Lincoln, iUniverse.
Jahn R., Dunne B., Dobyns Y., Nelson R., Bradish, G., Lettieri A., et al. (2000).
Mind/machine interaction consortium: PortREG replication experiments.
Journal of Scientific Exploration, 14, 499-555.
Knott K. (1975). A lunar signature in the geomagnetic ap-index. Earth, Moon, and
Planets, 14, 49-57.
May E.C., Utts J.M., and Spottiswoode J.P. (1995). Decision Augmentation Theory:
Toward a Model of Anomalous Mental Phenomena. Journal of Parapsychology,
59, 195-220.
O'Connor R.P., and Persinger M.A. (1999). Geophysical variables and behavior:
LXXXV. Sudden infant death, bands of geomagnetic activity, and pc1 (0.2 to 5
HZ) geomagnetic micropulsations. Perceptual and Motor Skills, 88, 391-397.
Palmer J., Simmonds C.A., and Baumann S. (2006). Geomagnetic fields and the
relationship between human intentionality and the hemolysis of red blood
cells. Journal of Parapsychology, 70, 275-301.
Persinger M.A. (1987). Geopsychology and geopsychopathology: mental processes
and disorders associated with geochemical and geophysical factors. Experientia,
43, 92-104.
171
Etzold
Radin D.I. (1997). The Conscious Universe. The scientific truth of psychic Phenomena.
San Francisco, HarperEdge.
Radin D.I., and Rebman J.M. (1998). Seeking Psi in the Casino. Journal of the
American Society for Psychical Research, 62, 193-219.
Radin D.I., Nelson R., Dobyns Y., and Houtkooper J. (2006). Reexamining
Psychokinesis: Comment on Bösch, Steinkamp, and Boller. Psychological
Bulletin, 132, 529-532.
Schmidt H. (1976). PK effect on pre-recorded targets. Journal of the American Society
for Psychical Research, 70, 267-291.
Schmidt H. (1981). Can an effect precede its cause? Foundations of Physics, 8,
463-480.
Schubert G., Sonett C.P., Smith B.F., Colburn D.S., and Schwartz K. (1975). Using
the Moon to probe the geomagnetic tail lobe plasma. Geophysical Research
Letters, 2, 277-280.
Spottiswoode S.J.P. (1990). Geomagnetic activity and anomalous cognition: A
preliminary report of new evidence. Subtle Energies, 1, 91-102.
Sturrock P.A. (2004). Time-Series Analysis of a Catalog of UFO Events: Evidence of
a Local-Sidereal-Time Modulation. Journal of Scientific Exploration, 18, 399–419.
Sturrock P., and Spottiswoode S.J.P. (2007). Time-series power spectrum analysis
of performance in free response anomalous cognition experiments. Journal of
Scientific Exploration, 21, 47-66.
Tsyganenko N.A. (1995). Modeling the Earth's magnetospheric magnetic field
confined within a realistic magnetopause. Journal of Geophysical Research, 100,
5599-5612.
Tsyganenko N.A., and Sitnov M.I. (2007). Magnetospheric configurations from a
high-resolution data-based magnetic field model. Journal of Geophysical Research,
112, A06225, doi:10.1029/2007JA012260.
Wilson B.W., Wright C.W., Morris J.E., Buschbom R.L., Brown D.P., Miller D.L., et
al. (1990). Evidence of an effect of ELF electromagnetic fields on human pineal
gland function. Journal of Pineal Research, 9, 259-269.
172
News Events, False Memory and Paranormal Belief
Neil Dagnall, Andrew Parker and Gary Munley

Department of Psychology & Social Change, Manchester Metropolitan University
Abstract
The current study explores Wilson and French’s (2006) finding

that believers in the paranormal are more susceptible to false
memories than non-believers. Seventy participants completed a
News Coverage Questionnaire concerning personal memories about
first seeing dramatic news events. The events included four news
items that are known to have been captured on film and two items
for which there is no known footage (the Princess Diana car crash
and the Bali Bombing). Presentation order was found to affect the
reporting of false memories. Endorsement of the Diana false
memory item positively correlated with scores on the Australian
Sheep-Goat Scale and the Dissocative Experience Scale when the
item appeared first but not when it appeared fifth. Believers in the
paranormal were more susceptible to false memory than non-
believers when the critical item was answered without first
recalling actual television footage. However, recalling previous
news footage facilitated endorsement of the false memory item
independent of paranormal belief and level of dissociativity.
Introduction
False memory has frequently been investigated using the

“crashing paradigm” (Crombag, Wagenaar, & van Koppen, 1996). This
involves examining participants’ willingness to report memories of a
highly emotional and consequential event that they could not have
witnessed (Crombag et al., 1996; Granhag, Strömwall, & Billings, 2003;
Ost, Vrij, Costall, & Bull, 2002; Wilson & French, 2006). This approach
Correspondence details: Neil Dagnall, Department of Psychology & Social Change, Manchester
Metropolitan University, Elizabeth Gaskell Campus, Hathersage Road, Manchester, M13 0JA
Email: N.Dagnall@mmu.ac.uk
173
was pioneered by Crombag, et al. (1996) and is based on the notion that
participants will be influenced by a critical question, which falsely
suggests that actual film footage exists (Crombag et al., 1996; Loftus,
1975).
Crombag et al. (1996) questioned participants 10 months after the
crash of an E1-A1 Boeing 747 into a block of flats in Amsterdam
(October, 1992). Although, there was no actual footage of the plane
crash the event received extensive media coverage. When asked, “Did
you see the television film of the moment the plane hit the apartment
building?”, 55% of the participants reported that they had seen non-
existent footage. In addition to this, 82% of participants provided
details about a non-existent fire. In a second study Crombag et al.
(1996) replicated these results. When participants were asked to
provide highly specific details about the crash (e.g., the angle at which
the plane crashed into the building) 66% of participants claimed to
have seen the footage and the majority were able to provide detailed
responses to the questions.
Other studies testing recollection of high profile events, for
which there is no known footage, have produced similar results: the
Diana, Princess of Wales, car crash (Ost, Vrij, Costall, & Bull, 2002); the
sinking of the Estonia ferry, in which almost 900 lives were lost
(Granhag et al., 2003); and the assassination of Dutch politician Pim
Fortuyn (Jelicic, Smeets, Peters, Candel, Horselenberg, & Merckelbach,
2006). Each of these studies produced impressive results. Ost et al.
(2002) using the Diana Princess of Wales car crash found that 44% of
participants were willing to report that they had seen actual television
footage of the crash. Granhag et al. (2003) noted that 38% of
participants in study 1, and 55% in study 2, reported seeing footage of
the sinking of the Estonia ferry. Jelicic et al. (2006) found that 63% of
participants claimed to have seen non-existent footage of the
assassination of Pim Fortuyn, although only 23% could provide details.
It is evident from these studies that participants will frequently report
having witnessed events they could not possibly have seen; events for
which no known video footage exists.
Wilson and French (2006) using a modified version of the
crashing memories paradigm examined the relationship between false
memory and paranormal belief/experience. They contended that
research in this area was important because variables associated with
susceptibility to false memories (e.g., dissociativity, fantasy proneness,
174
Dagnall, Parker & Munley
absorption) had also been found to correlate with paranormal
belief/experience (Wilson and French, 2006).
To test this, Wilson and French (2006) asked participants to
complete a News Coverage Questionnaire (NCQ), the Australian
Sheep–Goat Scale (ASGS) (Thalbourne, 1995), Dissociative Experiences
Scale (DES) (Bernstein & Putnam, 1986), and the Anomalous
Experiences Inventory (Kumar, Pekala, & Gallagher, 1994). The NCQ
contained four events that are known to have been captured on film
(the attack on the World Trade Centre, the Hillsborough football
stadium disaster, the toppling of the Saddam Hussein statue in Basra,
and the Challenger space shuttle disaster) and one item concerning
non-existent footage (the bombing of a nightclub in Bali). Participants
were asked specific questions about each of the events (where they
were, who they were with, and where they saw the footage). Questions
were also asked about the footage (e.g., was it in colour or black and
white, was there a commentary, and what was the picture quality like).
Wilson and French (2006) found that 36% of respondents falsely
claimed to have seen footage of the Bali bombing. Participants
reporting false memories were found to score higher on the ASGS, the
DES, and on the Belief, Experience and Ability subscales of the
Anomalous Experiences Inventory than participants who did not
report seeing the non-existent footage. On the basis of their findings
Wilson and French (2006) concluded that believers in the paranormal
may be more susceptible to false memories than non-believers. This
assertion, whilst logical, has yet to be firmly established and will be
examined within the current paper.
The present study
Whilst Wilson and French (2006) found support for the

hypothesis that believers in the paranormal may be more susceptible to
false memories than non-believers, other studies exploring the
relationship between susceptibility to false memories and the tendency
to report anomalous experiences have achieved only limited success
(French, 2003; French & Wilson, 2006). French, in his review of
eyewitness testimony and false memories for reports of anomalous
experiences (French, 2003; Wilson & French, 2006), suggests that this
may be because studies have used inappropriate techniques to
measure susceptibility to the “type” of false memories expected to
correlate with the tendency to report anomalous experiences.
175
Particularly, French suggests that techniques attempting to produce
detailed false memories for entire episodes (e.g., Loftus & Pickrell,
1995) may be more successful than techniques which attempt to
produce false memories for individual words (e.g., Roediger &
McDermott, 1995; word lists). This notion was supported by Wilson
and French’s (2006) findings with news events.
The problem with the Wilson and French (2006) study is that
endorsement of the false memory item may be conflated by the prior
exposure of participants to news events for which footage exists. In the
Wilson and French (2006) paper the critical Bali bombing event was
presented after two events for which there was actual footage.
Presenting the critical scenario in this way may encourage participants
to make reporting errors. Remembering actual footage may bias
participants’ responses when faced with an event for which no footage
exists; participants may assume the event was shown on television, or
that their memory of the event was based on television coverage. If
participants are being influenced in this way, it would be predicted
that item order would affect the reporting of false memories. Under
conditions where participants are first asked to recall real footage it
seems plausible that they will be more likely to misreport recalling
non-existent news footage. Indeed, Wilson and French (2006)
hypothesized that this may be an important factor requiring further
research. Noting that the level of false memory endorsement was
higher in their study than comparative studies (e.g., Jelicic et al., 2006),
they speculated that this was because previous studies used a single
event; Wilson and French (2006) presented the false memory amongst
four other events that had been filmed. For this reason the current
study assesses whether recalling footage of actual events influences
participants’ false memory responses.
This is achieved in the present paper by manipulating news
event order. The critical event will be presented first and last (fifth)
within an adapted version of the NCQ. The story selected for this
purpose is the fatal Paris car crash involving Diana, the Princess of
Wales, and her companion Dodi Fayed. This item has been used in
other similar studies, where participants have been found to wrongly
report seeing actual crash footage (Ost et al., 2002). This scenario was
selected over the Bali bombing because it received extensive coverage
and the story has remained in the news since the episode occurred. The
Bali bombing, whilst an important happening, has received less
prominent coverage since the event, consequently, it was felt that
176
participants would be less familiar with the Bali bombing. In order to
allow direct comparison with Wilson and French (2006), the Bali
bombing will still be used in the current study; it is presented at the
end of the experiment to determine if levels of false memory are
similar to those reported by Wilson and French.
Wilson and French (2006) hypothesized that item order would
influence the reporting of false memory. In particular they claimed that
recalling real news footage in preceding items would prime or activate
a generic schema for news coverage. Once activated, this news schema
influences the encoding and comprehension of non-existent news items
and leads to that information appearing more real. Under such
conditions, the individual is more likely to erroneously assert that such
footage actually existed. Thus it is predicted that presenting the Diana
false memory item last rather than first will increase false memory item
endorsement and that presenting the Bali bombing question at the end
of the current study will result in a higher level of endorsement than
that reported in the Wilson and French (2006) study. In addition to this
the present study will assess whether item order affects the
relationship between paranormal belief, dissociative experience and
false memory, In line with Wilson and French (2006) it is predicted that
level of paranormal belief and dissociative experience will be
associated with false memory endorsement and the rating of false
memory characteristics. The major contribution of the present study
will be to assess whether the relationship between paranormal belief,
dissociative experience and false memory is qualified by item order.
Method
Participants
Seventy respondents participated in this study. There were 25

males and 45 females. The mean age was 30.63 years (SD = 12.51) with
a range of 16–67 years. Participants included undergraduates and
employees from the Manchester Metropolitan University and
volunteers from the wider community.
Materials
Participants were asked to complete in the following order: an

adapted version of the News Coverage Questionnaire (NCQ) (Wilson
177
& French, 2006); the Australian Sheep–Goat Scale (ASG) (Thalbourne,
1995); and the Dissociative Experiences Scale (DES) (Bernstein &
Putnam, 1986). Questionnaire order was the same for all participants.
The News Coverage Questionnaire was designed by Wilson and
French (2006). The questionnaire contains five news items; four events
that are known to have been captured on film (the 2001 attack on the
World Trade Centre; the 1989 Hillsborough football stadium disaster;
the 2003 toppling of the Saddam Hussein statue in Basra; and the 1986
Challenger space shuttle disaster) and one item concerning non-
existent footage (the 2002 Bali nightclub bombing). Participants were
asked to indicate: whether they recall first seeing the event; where they
saw the event; and who they were with at the time. Questions were
asked about the actual news item footage: whether the clip was in
colour or black and white; whether there was an audible commentary
on the footage; and how participants would rate the picture quality.
These questions were intended to ensure that participants reporting
false memories based their report upon a specific memory of viewing,
rather than just recalling the circumstances under which they first
heard the news being reported.
In the current study an additional item was added. This
concerned watching non-existent footage of the 1997 car crash
involving Princess Diana. This item was selected because participants
have been found to often report seeing the actual video footage of the
crash in which Diana, Princess of Wales, and Dodi Fayed died, even
though no footage exists (Ost et al., 2002). This news event shares
many important features with the El Al air crash used in Crombag et
al.’s (1996) study: the media extensively covered the story; television
footage appeared soon after the event; the crash was the main news
story; the story was told again and again illustrating what footage was
actually available; people knew many details about the event without
actually seeing it; and a simulation of the event was shown on some
news stations.
In order to allow the current study to assess the effect of item
order on false memory, the order of the items for which there is no
known television footage was manipulated. The Diana news event
appeared in either first or fifth; that was before or after the events for
which there was known footage (World Trade Centre, Hillsborough,
Space Shuttle Challenger and Basra). The second critical item, the Bali
Bombing was presented at the end of the study to allow direct
178
comparison with Wilson and French (2006). The NCQ was followed by
the Australian Sheep–Goat Scale (ASG) and DES.
The Australian Sheep–Goat Scale (ASGS) measures belief in, and
alleged experience of, the paranormal. It consists of 18 items that relate
to three core concepts of parapsychology: life after death,
psychokinesis, and extrasensory perception. The response options are:
False (scored as zero), “?” (Don’t know: scored as 1 point), and True
(scored as 2 points). The scale has a range from zero to 36, higher
scores indicating higher levels of belief and experience. The ASGS has
proven reliability and validity (Thalbourne, 1995; Thalbourne and
Delin, 1993).
The Dissociative Experiences Scale (DES; Bernstein & Putnam,
1986) is a 28-item self-report measure that assesses the frequency of
dissociative experiences on a response scale that starts at 0% and
increases by 10% increments up to 100% (Carlson & Putnam, 1993).
Responses across all items are averaged to obtain a mean DES score
(range 0–100), with higher scores indicating higher levels of
dissociative symptoms.
Procedure
Participants were informed that the current study was

investigating memory for news events. Following completion of the
questionnaire participants were debriefed. They were informed that
there was no actual footage of the Diana crash and the Bali bombing.
Results
Both the Australian Sheep–Goat Scale (ASGS) (α = .92) and

Dissociative Experiences Scale (DES) (α =.90) were found to have
excellent internal reliability. Of the 70 participants who completed the
questionnaire, 68 (97%) reported that they had seen the video footage
of the attack on the World Trade Centre, 26 (37%) reported having seen
the video footage of the Hillsborough football disaster in 1989, 65 (93%)
recalled video footage of the statue of Saddam Hussein being torn
down in Basra in 2003, and 22 (31%) recalled the video footage of the
Challenger space shuttle disaster in 1986. Regarding the non-existent
video footage, 57 (81%) participants claimed to have seen the car crash
involving Princess Diana; endorsement was higher on this item than
on two of the events for which actual video footage exists
179
(Hillsborough and the space shuttle Challenger). Table 1 contains the
percentage claiming recall for the present study and for Wilson and
French (2006).
Table 1: Percentage Claiming to Recall Television Footage of Specific News Events in the Current
Study and in the Wilson and French (2006) Study
Percentage claiming recall

News story Current study Wilson & French (2006)
World Trade Center 97 97
Hillsborough 37 30
Basra 93 72
Challenger 31 40
Bali 53 36
Diana 81 NA
To determine whether endorsement of the Diana crash (false

memory item) was influenced by scenario presentation order
(presented first vs. fifth) chi-square was used. A significant association
was found between presentation order and item endorsement, χ2 =
5.14, df = 1, p = .023, Cohen’s d = 0.56. In particular, 77% of correct
rejections occurred when the Diana event was presented first this fell
to 23% when the event was presented fifth (last), χ2 = 3.77, df = 1, p = .
052, Cohen’s d = 0.54. Endorsement of the Diana crash was further
studied by examining false alarms, the number of incorrect
endorsements produced, This found that presentation order had no
significant effect on the number of false alarms produced, χ2 = 1.42, df
=1. p > .05, Cohen’s d = 0.32; the trend towards fewer false alarms 42%
when presented first, compared to 58% when presented fifth was,
however, in the predicted direction. Finally, the difference between
correct rejections and false alarms as a function of presentation order
was analysed. True memory was found to be higher when the Diana
crash was presented first than when it was presented fifth, χ2 = 5.82, df
=1. p = .016, Cohen’s d = 0.78.
Of the 57 participants who claimed to have seen non-existent
footage of the Diana crash, 88% were willing to state where they were
when they saw the footage, 71% recalled who they were with at the
time, and 51% recalled what TV channel/website they saw the footage
on (95% claiming it was in colour). The picture quality of the footage,
on a 7-point scale was 5.12 (SD = 1.47) and the mean clarity of memory
was 5.00 (SD = 1.82).
180
The mean clarity rating for events with actual footage was
compared with the mean rating for the Diana non-existent television
footage. Participants who endorsed having seen the Diana footage (M =
5.00) claimed to recall it as clearly as they did actual footage (M = 5.02);
t (56) = - 0.09, p > .05.
A median split was performed on ASGS scores to produce high
(above median) and low (below median) conditions. A 2 (Level of
Paranormal Belief: high vs. low) x 2 (Presentation Order: first, fifth)
factor between subjects ANOVA was then conducted to determine
whether variable level (high vs. low) and presentation order affected
participants’ clarity ratings of the non-existent Diana car crash footage
(see Table 2).
Table 2: Clarity Ratings of the Non-Existent Diana Car Crash Footage as a Function of level of
Paranormal Belief
Level of paranormal belief

Below Median Above Median Overall
Presentation order M SD M SD M SD
First 4.80 1.62 4.93 1.33 4.88 1.42
Fifth 3.86 2.18 6.00 1.49 5.09 2.08
Overall 4.25 1.94 5.55 1.50
A significant main effect was noted for Level of Paranormal

Belief, F(1, 53) = 6.24, p = .016, ηp2 = 0.11. Participants high in
paranormal belief (M = 5.55, SD = 1.50) rated the clarity of the non-
existent footage higher than participants low in paranormal belief (M =
4.25, SD = 1.99).
No significant effect was found for Presentation Order, F(1, 53)
= .20, p > .05. The interaction between Level of Paranormal Belief and
Presentation Order, F(1, 53) = 4.91, p = .031, ηp2 = 0.09 was significant.
Simple main effect analysis on each level of Presentation Order
showed that Level of Paranormal Belief had no effect on clarity rating
when the Diana car crash question was presented first, t(22) = 0.21, p
> .05, however, it did affect ratings when the Diana car crash question
was presented fifth. Participants high in paranormal belief (M = 6.00,
SD = 1.48) were found to rate clarity for the non-existent footage higher
than those low in paranormal belief (M = 3.86, SD = 2.18; t(31) = 3.36, p
= .002, Cohen’s d = 1.21).
A median split was performed on DES scores to produce high
(above median) and low (below median) conditions. A 2 (Level of
181
Dissociativity: high vs. low) x 2 (Presentation Order: first, fifth)
ANOVA was then conducted to determine whether DES score (high
vs. low) and presentation order affected participants’ clarity ratings of
the non-existent Diana crash footage (see Table 3).
Table 3: Clarity Ratings of the Non-Existent Diana Car Crash Footage

as a Function of level of Dissociativity
Level of dissociativity
Below Median Above Median Overall
Presentation order M SD M SD M SD
First 5.18 1.66 4.61 1.19 4.88 1.42
Fifth 4.31 2.32 5.60 1.79 5.09 2.08
Overall 4.71 2.05 5.21 1.64
No significant main effect was found for Level of Dissociativity

F(1, 53) = 0.56, p > .05. No significant main effect was noted for
Presentation Order, F(1, 53) = 0.13, p > .05. The interaction between
Level of Dissociativity and Presentation Order F(1, 53) = 3.65, p > .05
was not significant.
A series of correlations were conducted to investigate the
relationship between Australian Sheep–Goat Scale (ASGS) and
Dissociative Experiences Scale (DES) scores, and whether or not
participants claimed to have seen television footage of the car crash
involving Princess Diana. ASGS was found to positively correlate with
participants’ endorsement of the non-existent Diana television footage
(r = .38, n = 70, p = .001). A significant correlation was found between
ASGS and the clarity rating of the non existent Diana memory (r = .35,
n = 57, p = .007). DES was found to positively correlate with
participants’ endorsement of the non-existent Diana television footage
(r = .28, n = 70, p = .017). No significant relationship was found between
DES and the clarity rating of the non existent Diana memory (r = .17, n
= 57, p > .05). The DES and ASGS were also found to correlate (r = .37, n
= 70, p = .002).
Further correlations were conducted on participants’ responses
when the Diana item was presented first and fifth. When the Diana
item appeared first positive correlations were found between false
memory endorsement and paranormal belief (r = .43, n = 34, p = .011)
and between false memory endorsement and DES (r = .39, n = 34, p = .
022). When the Diana item appeared fifth no significant relationship
was found between false memory endorsement and paranormal belief
182
(r = .28, n = 36, p > .05) and between false memory endorsement and
DES (r = .16, n = 36, p > .05).
These results indicate a significant positive correlation between
false memory and belief in the paranormal and DES when the false
memory item is not preceded by events for which footage exists.
However, no relationship was found between false memory and belief
in the paranormal and DES when the false memory item was presented
after events for which there was known footage.
Endorsement of the non-existent Bali bombing footage was
presented at the end of the questionnaire in the current study to
investigate whether presentation order affected the proportion of
correct and incorrect responses. The data from this paper, where the
event was presented last was compared with that of Wilson and
French (2006); Wilson and French (2006) always presented the Bali item
third of five. As predicted presenting the Bali item last produced a
higher percentage of false endorsements (53% compared to 36% in the
Wilson & French, 2006 study), χ2 = 4.77, df = 1, p = .029, Cohen’s d =
0.54. Endorsement of the Bali false memory item may have been higher
in the current study than the French and Wilson (2006) study because
participants were influenced by previously recalling actual footage
(World Trade Centre, Hillsborough, Basra, etc.)1.
Discussion
Consistent with previous studies a substantial proportion of

participants erroneously reported seeing non-existent news footage of
critical events (crash of Diana and Bali bombing) (Crombag et al.,1996;
Granhag et al., 2003; Ost et al., 2002; Smeets et al., 2006; Wilson &
French, 2006). The current paper investigated the hypothesis that the
reporting of false memories would be affected by presentation order as
predicted by Wilson and French (2006). It was expected that a higher
proportion of participants would report seeing the Diana crash when
the critical question was presented fifth rather than first, and that
presenting the Bali bombing question at the end of the current study
would result in a higher level of endorsement than that reported in the
Wilson and French (2006) study. Both of these predictions were
supported: 77% of correct rejections occurred when the Diana item was
presented first; and higher levels of endorsement were produced for
1
Whilst it is unusual to combine data in this way, studies in this area frequently use percentage
comparisons to support/refute points (e.g., Smeets, et al., 2006; Wilson & French, 2006).
183
the Bali item than were found in the Wilson & French (2006) study
(53% vs. 36%).
The influence of serial order is explained by Wilson and French
(2006) as due to the activation of a generic news schema. Presumably,
once primed, the news schema provides a basis for the encoding and
comprehension of incoming information. Non-existent footage is thus
assimilated into this schema. When probed about the details of this
footage, the generic schema is reactivated and false memories of non-
existent events are the result. Interestingly, clarity ratings also
increased when the critical item was presented fifth, at least for those
high in paranormal belief. It would appear that these individuals had
more vivid false recollections of the details of the non-existent event.
Exactly why this was found in terms of the news schema
hypothesis is difficult to say. Schematic information, by its very nature,
is abstracted away from specific perceptual or episodic details that may
influence ratings of clarity (Alba & Hasher, 1983). Indeed, memory for
schematic information is less likely to be associated with clear and
vivid recollections (Lampinen, Faires, Neuschatz, & Toglia, 2000). If
this is true, then the news schema hypothesis of Wilson and French
(2006) may not provide a full, account of the effects observed here.
Instead, perhaps one explanation is that individuals high in
paranormal belief are more prone to incorporating perceptual or
episodic features form real events into their false memories. In other
words, they are more likely to “import” details from studied events
into non-studied events making them seem more likely and real (e.g.,
Lyle & Johnson, 2006). Another possibility is that they adopt a more
liberal criterion for assessing the degree of clarity and detail retrieved.
Both these ideas warrant further investigation.
Previous research has also found that those score high on
measures of paranormal belief also demonstrate reality monitoring
deficits (Irwin, 2004). As false memories can also arise when
individuals fail to accurately engage in reality monitoring, then false
recollections of the Diana crash may also be due to reality monitoring
failures in those endorsing high levels of paranormal belief.
The results of the current study indicate that participants are
more likely to wrongly report seeing television coverage of an event
when the critical false memory item is preceded by events for which
actual footage exists. This supports the notion that remembering actual
footage biases participants’ responses; participants may assume the
“critical event” was shown on TV, or that their memory of the event
184
was based on television coverage (Wilson & French, 2006).
Participants scoring above the median on the ASGS were found to rate
the clarity of the non-existent Diana footage higher than those scoring
below the median on the ASGS when the Diana question was
presented fifth. No difference was evident when the Diana question
was presented first.
In the current study overall endorsement of the non-existent
Diana television footage was higher than that reported by Ost et al.
(2002); 81% vs. 44%. This may be partially explained by the use of
Wilson and French’s (2006) multi-event paradigm, but may also be due
to other factors: the continued news coverage since the event; recent
refocusing (Stevens Inquiry, 2006); and the forthcoming tenth
anniversary coverage (August 2007). This notion is consistent with the
imagination inflation effect (Garry et al., 1996). Whereby, intense and
sustained media coverage is likely to encourage people to imagine
details of an event and motivate the belief that footage has appeared on
television.
The overall clarity rating for false memory in the current study
was moderately high similar to that produced by Wilson and French
(2006) (5.00 compared to 4.25: on a 7-point scale). As in the Wilson and
French (2006) study the lowest clarity ratings were produced for the
Hillsborough and Challenger disasters. This is perhaps not surprising
as these are the two oldest events. Interestingly, for participants who
“wrongly” recalled seeing the Diana crash there was no difference
between their false memory clarity rating and that for the real events
(5.00 vs. 5.02). This finding is in line with previous research (e.g.,
Porter, Yuille, & Lehman, 1999) that has demonstrated that fabricated
memories are sometimes rated as vivid and detailed as real memories
(Smeets et al., 2006). In addition to this, a high proportion of
participants provided details to accompany their endorsement of the
Diana item.
In support of Wilson and French (2006), the current study found
positive correlations between endorsement of the Diana item and
scores on the Australian Sheep–Goat Scale (ASGS) and Dissociative
Experiences Scale (DES). These findings are consistent with previous
studies, which have reported relationships between false memory and
psychological correlates of paranormal belief (e.g., fantasy proneness:
Irwin, 1990; hypnotic suggestibility: Kumar & Pekala, 2001; absorption:
Irwin, 1985; dissociativity: Wolfradt, 1997) (French & Wilson, 2006).
185
A major finding in the current study was that the relationship
between false memory and scores on the AGS and DES was qualified
by item order. A medium positive correlation was found between
belief in the paranormal (ASGS), DES, and false memory for
“spontaneous” endorsement of the Diana false memory item.
However, this relationship was weakened when endorsement was
induced by thinking about actual television footage of news events.
The fact that endorsement of the false memory item was affected by the
presentation of preceding events, for which there is known footage,
suggests that recalling actual footage influenced responses on the
critical Diana item independent of paranormal belief and dissociativity.
This finding is consistent with Smeets et al. (2006), who
demonstrated that reporting of non-existent footage was shaped by
misleading and/or ambiguous questions, and that the endorsement
rate of footage depends on the way in which participants are
questioned. Further to this, they identified the suggestibility of the
participant as being important (Smeets et al., 2006). Similarly, studies
have demonstrated that verbal suggestion can affect participants’
perception of paranormal events, such as séances (Wiseman, Greening
& Smith, 2003) and key bending (Wiseman & Greening, 2005). Like
these previous studies the current study is limited by the fact that it
fails to explain the relationship between false memory, paranormal and
psychological variables; it is unclear whether the questions affect
perception, memory, or whether participants are influenced by
demand characteristics (French & Wilson, 2006).
The use of the crashing paradigm to assess false memory is not
without criticism. The main problem being that it does not necessarily
follow that participants claiming to have seen non-existent news
footage have actually created false memories (Smeets, Merckelbach,
Horselenberg, & Jelicic, 2005). Statement endorsement may be
influenced by myriad cognitive and social factors (Smeets et al., 2006):
participants may provide socially desirable answers; they may
misinterpret the critical questions (refer to the aftermath rather than
the actual event); use general knowledge heuristics; or make source
monitoring errors rather than create false memories. Subsequent
studies are required to assess the reasons for false memory
endorsement.
In addition to this the current multi-event paradigm has the
problem of being unable to ensure that memory for actual footage is
accurate. This is not important to the degree that the current study is
186
exploring whether participants endorse having seen non-existent
footage. However, the fact that memories for events for which there is
known footage may be untrue does bring into question the validity of
subjective ratings. One obvious way of overcoming this problem is to
show participants a series of news clips and then after delay, ask them
about the events alongside events for which there is no known footage.
This modification, whilst time consuming would ensure that
participants had been exposed to the footage and that “true” memories
had a clearly identifiable objective basis.
References
Alba J., & Hasher, L. (1983). Is memory schematic? Psychological Bulletin, 93, 201–231.
Bernstein, E. M., & Putnam, F. W. (1986). Development, reliability and validity of a
dissociation scale. Journal of Nervous and Mental Diseases, 74, 727–735.
Carlson, E. B., & Putnam, F. W. (1993). An update on the dissociative experiences
scale. Dissociation, 6, 16-27.
Crombag, H. F. M., Wagenaar, W. A., & van Koppen, P. J. (1996). Crashing memories
and the problem of source monitoring . Applied Cognitive Psychology, 10,
95-104.
French, C. C. (2003). Fantastic memories: The relevance of research into eyewitness
testimony and false memories for reports of anomalous experiences. Journal of
Consciousness Studies, 10, 153-174.
French, C. C., & Wilson, K. (2006). Incredible memories: How accurate are reports of
anomalous events? European Journal of Parapsychology, 21, 166-181.
Garry, M., Manning, C. G., Loftus, E. F., & Sherman, S. J. (1996). Imagination inflation:
imagining a childhood event inflates confidence that it occurred. Psychological
Bulletin and Review, 3, 208–214.
Granhag, P. A., Strömwall, L., & Billings, J. (2003). I'll never forget the sinking ferry:
how social influence makes false memories surface. In Vanderhallen, M.,
Vervaeke, G., J. van Koppen, P., & Goethals, J. (eds.). Much ado about crime:
Chapters on psychology and law. Belgium: Uitgeverij Politeia, 129-140.
Irwin, H., J. (1985). Parapsychological phenomena and the absorption domain. Journal
of the American Society for Psychical Research, 79, 1-11.
Irwin, H., J. (1990). Fantasy proneness and paranormal beliefs. Psychological Reports,
66, 655-658.
Jelicic, M., Smeets, T., Peters, M. J. V., Candel, I., Horselenberg, R., & Merckelbach, H.
(2006). Assassination of a controversial politician: remembering details from
another non-existent film. Applied Cognitive Psychology, 20, 591-596.
Kumar, V .K., & Pekala, R. J. (2001). Relation of hypnosis-specific attitudes and
behaviors to paranormal beliefs and experiences: a technical review. In
Houran, J., & Lange, R. (eds.). Haunting and poltergeists: multidisciplinary
perspectives. Jefferson, NC: McFarland & Co, 260–279.
187
Kumar, V. K., Pekala, R. J., & Gallagher, C. (1994). The Anomalous Experiences
Inventory. Unpublished psychological test. West Chester, PA: West Chester
University.
Lampinen, J. M., Faires, J. M., Neuschatz, J. S., & Toglia, M. P. (2000). Recollections
of things schematic: the influence of scripts on recollective experience. Applied
Cognitive Psychology, 14, 543-554.
Loftus, E. F. (1975). Leading questions and eyewitness reports. Cognitive Psychology, 7,
560-572.
Loftus, E. F., & Pickrell, J. E. (1995). The formation of false memories. Psychiatric
Annals, 25, 720-725.
Lyle, K. B., & Johnson, M. K. (2006). Importing perceived features into false
memories. Memory, 14, 197-213.
Memon, A., & Young, M. (1997). Desperately seeking evidence: The recovered
memory debate. Legal and Criminological Psychology, 2, 131-154.
Ost, J., Vrij A., Costall A., & Bull, R. (2002).Crashing memories and reality monitoring:
distinguishing between perceptions, imaginings and false memories. Applied
Cognitive Psychology, 16, 125–134.
Porter, S., Yuille, J.C., & Lehman, D. (1999). The nature of real, implanted, and
fabricated memories for emotional childhood events: Implications for the
recovered memory debate. Law and Human Behavior, 23, 517-538.
Pekala, R. J., Kumar, V. K., & Marcano, G. (1995). Anomalous/paranormal
experiences, hypnotic susceptibility, and dissociation. Journal of the American
Society for Psychical Research, 89, 313–332.
Roediger, H. L. III, & McDermott, K. B. (1995). Creating false memories:
Remembering words not presented in lists. Journal of Experimental
Psychology: Learning, Memory, and Cognition, 21, 803-814.
Smeets, T., Jelicic, M., Peters, M. J. V., Candel, I., Horselenberg, R., & Merckelbach, H.
(2006). ‘Of course I remember seeing that film?’ - How ambiguous questions
generate crashing memories. Applied Cognitive Psychology, 20, 779-789.
Smeets, T., Merckelbach, H., Horselenberg, R., & Jelicic, M. (2005). Trying to recollect
past events: confidence, beliefs, and memories. Clinical Psychology Review, 25,
917-934.
Thalbourne, M. A. (1995). Further studies of the measurement and correlates of belief
in the paranormal. Journal of the American Society for Psychical Research, 89, 233–
247.
Thalbourne M. A., & Delin, P.S., (1993). A new instrument for measuring the sheep–
goat variable: its psychometric properties and factor structure, Journal of the
Society for Psychical Research, 59, 172–186.
Wilson, K., & French, C.C. (2006). The relationship between susceptibility to false
memories, dissociativity, and paranormal belief and experience. Personality and
Individual Differences, 41, 1493-1502.
Wiseman, R. & Greening, E. (2005). 'It’s still bending': Verbal suggestion and alleged
psychokinetic metal bending. British Journal of Psychology, 96, 115-127.
Wiseman, R., Greening, E., & Smith, M. (2003). Belief in the paranormal and
suggestion in the séance room. British Journal of Psychology, 94, 285-297.
Wolfradt, U. (1997). Dissociative experiences, trait anxiety, and paranormal beliefs.
Personality and Individual Differences, 23, 15–19.
188
Book Review
Matt Colborn
Independent Researcher, UK
An Essay Review of “Irreducible mind: Towards a psychology

for the 21st Century” Edited by Edward F. Kelly and
Emily Williams Kelly
(2006)
Irreducible mind: Towards a psychology for the 21st century is a

radical book for an epoch of extreme neurological reductionism. The
causes of behaviours, personality and consciousness are increasingly
sought in localised brain areas and genes, and pretty much nowhere
else; 'there is nothing more, no magic, no additional components to
account for every thought, each perception and emotion, all our
memories, our personality, fears, loves and curiosities. (O'Shea, 2008, p.
12).' Accompanying this is a widespread attempt to erase views of the
mind that are 'occult' or 'immeasurable' (Dennett, 1978; Rorty 1979).
The authors of Irreducible Mind (henceforth IM) believe that
there is more than enough evidence to demonstrate that this dominant
view is at best incomplete and at worst false. The central problem is the
relation between the private subjective 'first person' inner world and
the outer, objective 'third person' world of physiological events and
processes in the body and brain (p. xvii, E.F. Kelly).
A far more inclusive theory is also needed to accommodate the
full range of human experiences, which includes 'rogue phenomena'
like psi effects, mind-body interaction, mystic experience and others.
F.W.H. Myers sketched the foundations of such a theory over a century
ago in his Human Personality and its Survival of Bodily Death (1902),
which is included on an accompanying CD. In Human Personality,
Myers developed a theory of the 'Subliminal Self.' This was the idea
that:
Correspondence details: Matt Colborn, 14 Hall Road, Haconby, Bourne, Lincolnshire, PE10 0UY,
United Kingdom. Email: Matt.Colborn@googlemail.com
189
Contacts by Distressed Individuals to UK Academic Research Units
There exists a more comprehensive consciousness… from which the

consciousness and faculty of earth-life are mere selections…. [N]o Self of
which we can here have cognisance is in reality more than a fragment of a
larger Self – revealed in a fashion at once shifting and limited through an
organism not so framed as to afford it full manifestation. (Myers, 1902, v2,
pp12-15. Quoted in IM p.73)
Myers' Subliminal Mind contained the potential for genius as

well as pathology and the emergence of what he termed 'supernormal
phenomena' – things that apparently belonged to a more advanced
stage of evolution (Myers, 1885). In addition, he thought that the brain
should be viewed as a 'filter' that limited consciousness rather than its
producer. Myers' theory is therefore strongly dualistic. Most
contemporary researchers reject any form of dualism out of hand
(Dennett, 1991), but this has resulted in the toleration of some very
unsatisfactory models of mind (Searle, 1992).
IM in part presents the case for some kind of interactive dualism.
It begins with an overview of contemporary cognitive neuroscience
(Chapter 1) and the historical foundations of an alternative approach
(chapter 2) before tackling problems of the mind's influence on the
body (chapter 3), memory (chapter 4), automatism and secondary
centres of consciousness (chapter 5), unusual experiences near death
(chapter 6), genius (chapter 7) and mystical experience (chapter 8).
Chapter 9 is a re-assessment of Myers' theory and an attempt to
provide a sketch of an 'expanded psychology'. For clarity, the
individual chapter reviews are subtitled.
Chapter 1: A view from the mainstream: contemporary cognitive

neuroscience and consciousness debates. [E.F. Kelly].
This chapter provides a potted history of cognitive neuroscience

over the twentieth century, starting with behaviourism and proceeding
to the cognitive revolution. This “can be characterized… as a
movement toward progressively less unsatisfactory analyses of the
mind” (p. 2).
Behaviourism ignored the inner workings of the mind in favour
of observable behaviour; complex behaviours were believed to be
combinations of simpler stimulus-response reactions. The behaviourist
programme gave way in the 1950s and 1960s to a family of positions
190
Colborn
known collectively as identity theory. Identity theory states that

associated mental and physical states are in some sense identical (Feigl,
1958; Putnam, 1967). Putnam (1967) favoured a variety of identity
theory called functionalism, which is promoted by contemporary
writers like Dennett (1991). Functionalism defines mental states not in
terms of what they are made of, but what they do. Just as 'cutting tools'
might be made of stone, metal or laser beams, so mental states might
be realizable in either brains or computers provided they serve similar
functions.
Computers were central to the cognitive revolution because of
the idea that a sophisticated computer or Artifical Intelligence (AI)
might perform cognitive functions comparable to that of people
(Turing, 1950). Some philosophers rejected the idea that a digital
computer could ever be conscious. Searle (1992, 1980) argued that since
computation was just symbol shuffling, even a sophisticated program
could never, for instance, understand language in the way humans do.
However, Searle (1984) also advocates a form of 'biological naturalism'
(brains = mind), a stance Kelly rejects. I personally find Searle's
arguments inconclusive, and they are of course regularly pilloried by
writers of a more computationalist bent (e.g. Hofstader, 2007).
The 'second' cognitive revolution involved connectionism and
dynamic systems, both of which try to address the inadequacies of
classical AI. These newer approaches still rely on simulation on a
digital computer, with its ultimate limitations.
There is currently significant dissatisfaction with dominant
theories of mind, which belies the triumphal claims of writers like
Pinker (2004). Chomsky (1993), a pioneer of the cognitive revolution,
complained that it was premature to try and reduce psychology to
present-day neurophysiology, and others have made similar noises.
Despite the hype, there is a clear need to move beyond the mainstream
to develop new mind-models.
Chapter 2: F.W.H Myers and the empirical study of the mind-body

problem [E.W.Kelly]
George Santayana's comment (1905) about those who forget

history being doomed to repeat it is certainly true of psychology. Many
of the current claims that the 'self,' 'free will,' 'qualia,' etc. are illusions
191
are actually just the latest versions of assertions made in the nineteenth
century.
In the latter half of the nineteenth century, psychology was
transformed from a branch of moral philosophy to a naturalistic
science. At the time, mechanistic determinism – the notion of the
Universe as machine that obeyed laws -- was the dominant paradigm
in science. A result of this was the emergence of epiphenomenalism,
the view that mind and free will are illusions or at best secondary,
ineffectual by-products of the physical, mechanical world. The idea
that the mind might be a fundamental, causal agent in the universe
was rejected from the start because this was seen as a primitive,
'supernatural' way of thinking.
However, this rejection had a cost, which was a conflict between
experience and knowledge. Individual first-person experience suggests
one kind of world of personal agency (free will, volition), but science
suggested another kind of world, that of impersonal agency only. This
conflict led to a dilemma; either psychology should be expanded to
include subjective agency, or else they could narrow psychology to fit
science as it was then understood. Historically, the mainstream has
chosen the latter course.
Myers remained committed to the idea that the mind might be a
fundamental causal agent. This alone was unpalatable to many
contemporary scientists, and would be even more so today. However,
Myers had many important insights into the nature of consciousness,
which are based on many, many case studies. In many ways, he was
the Darwin of psychology, in that he was able to organise and classify
many different but overlapping phenomena into a coherent whole. The
latter half of this important chapter gives an overview of this
classification.
Chapter 3: Psychophysiological influence [E.W.Kelly]
The influence of the mind on the body's health, or otherwise, is

one of the great unknowns of medicine. This has been denied, of
course; the currently prevailing view is that 'the mind-body problem…
cannot be viewed as the subject matter of psychosomatic medicine
(Lipowski, 1984)! However, there are a wide range of phenomena that
beg important questions about the relationship between the body and
the mind. 'Psychophysiological influence' refers to a large range of
192
Colborn
phenomena, from the relatively conventional 'placebo effects' to the

hypnotic induction of wounds and marks on the skin (Moody, 1946,
1948), to very odd cases where previous lives seem to have left their
mark on the claimant's body (Stevenson, 1997).
The assessment of even relatively conventional phenomena like
the placebo effect is not easy. Collins and Pinch (2005) describe placebo
research as a 'hall of mirrors,' plagued with expectancy effects and
reporting biases. Similar problems emerge all over this field. One
running theme is that phenomena tend to get ignored until someone
conceives of a (conventional) mechanism to explain them. An example
of this is the curing of warts by suggestion, which is a relatively well-
attested occurrence. Most of the theories have proposed the alteration
of the blood-supply to the warts. Kelly argues that these explanations
must be incomplete because some cases involve, for example, the
selective reduction of warts.
As an outsider to these fields, I find it very difficult to assess just
how well established many of these phenomena really are. For
example, Buckman (2007) claimed that there was little persistent
evidence that mental attitude affected the course of cancer. Kelly cites
several small studies by Meares (1977, 1979,1980) that seem to show
that some cancer patients did benefit from an intensive course of
meditation – including five that had a complete remission. However
Coyne, Stefanek & Palmer (2007) recently published a study on head
cancer in which they claimed that psychotherapy had no effect. Emily
Kelly (personal communication) responded that there are many factors
involved, and ones that are not likely to be reproduced easily in studies
like Coyne, Stefanek & Palmer.
Despite these issues, if even only a small fraction of the cases
cited by Kelly are legitimate, then they show that our understanding of
the interrelationship between the mind and the body is not complete.
So although this field has problems, it is a very rich one that deserves
far more study than it gets.
Chapter 4: Memory [A.Gauld]
Alan Gauld's discussion of memory and the brain covers firstly

problems with what he refers to as the 'memory trace doctrine' and
secondly, issues relating to memory and the question of Survival of
bodily death.
193
There are several serious problems with the idea that our memories are
stored as physical 'traces' in the brain. William James' noted that the simple
revival of an image into the mind is not enough to make it a memory. Such an
image is 'a duplicate, a second event, having absolutely no connection with
the first event except that it happens to resemble it….' (James, 1890, pp 649—
50, cited in IM p.243).
Many explanations of the revival of a past memory presuppose

its function and cannot really explain memory itself. Other issues
include the richness and form of associations with a particular
memory, the literal nature or otherwise of 'mental images' and the
feeling of past-directness of personal memories. The latter dilemma is
connected to the problem of intentional recall, and general issues of
Intentionally. These issues are difficult, but Gauld's prose is precise
and deserves careful attention.
Contemporary research into memory tends to be technically
sophisticated but philosophically naïve. The Computational Theory of
Mind (CTM) assumes that memory works in a way somewhat
comparable to the information processing of computers. Many
psychologists still discuss memory in terms of 'input' 'encoded' by
successive stages of sensory pathways and later 'retrieved' and further
processed. (Gauld regards neural network and dynamic systems
models of memory as on the same footing as conventional digital-
computer models in that their discrete internal memory states are still
representations of a kind). Many of the conceptual issues remain.
Contemporary neuroscience is currently dominated by
NeuroImaging techniques like PET and fMRI scans. Rather than
providing clear answers to issues of memory, fMRI studies have
produced rich cohorts of new data, much complexity and many new
questions. These studies are also fraught with conceptual and
methodological problems including poor to nonexistent replication
(Harpaz, 1997). Gauld also notes that the early studies in particular led
to some 'breathtaking oversimplifications;' like claims of the 'God spot'
or 'seat of intelligence (p.267).'
Gauld concludes that many of these difficulties suggests that 'a
fuller understanding of human memory may ultimately require some
radical change of perspective.' (p. 281), especially if we take account of
the evidence for the survival of memories beyond death. Any theory
194
Colborn
will be need to be“top down” rather than “bottom up,” will

accommodate the fact that some declarative memories cannot be tied to
localities in the brain, and will regard nerve tracts not as conveyers of
information but as 'means by which spatiotemporal patterns of activity
in different regions may be fine-tuned to create overarching patterns
(p. 228).' Each of these points seems pretty reasonable to me, and
would begin address the naiveté of much current memory research.
Chapter 5: Automatism and secondary centres of consciousness

[Adam Crabtree]
'Automatism' refers to the occurrence in some people of 'a set of

memories, thoughts and feelings which are extra-marginal and outside
of the primary consciousness altogether (James 1902/1958. P. 188, cited
in IM, p.302). Dissociative Identity Disorder (DID, formerly known as
Multiple Personality Disorder), where a number of distinct
personalities appear to coexist within one person, is probably the most
familiar of these conditions, but there are a whole spectrum of
'dissociative' phenomena that can also be classed as automatisms.
These were crucial for the development of Myers' theory of the
'subliminal mind,' and he thought that the explanation for
automatisms needed to be psychological rather than physiological.
This was partly because the 'secondary centres of consciousness,' of
which the subject was apparently not consciously aware, often
appeared to exhibit high levels of volition and intelligence. Examples
are the messages sent by 'communicators' through automatic writing
and also that some 'secondary personalities' of DID suffers can appear
more intelligent and mature than their primary personalities.
The chapter goes on to discuss a number of other historical
theories that were held before the subject of automatisms was, like so
many other psychological phenomena, pushed into obscurity by the
rise of behaviourism. It's somewhat disheartening to read that more
recent theorists like Kihlstrom (1993) seem only just to be reaching
conclusions that were reached by James and Myers a century ago.
Volition sits at the heart of the contemporary debates on
automatisms, and these phenomena have been used to assert that
volition or free will is an illusion. Wegner (2002) rejects entirely the
195
idea that human beings are free agents, claiming that all behaviour is
involuntary. In his view, conscious will is simply an illusion.
In this interpretation, automatisms are simply behaviours that
occur without the illusory feeling that it's produced by conscious will.
Kelly observes that this ignores the close association of automatisms
with supernormal phenomena, creativity and genius. I'm not sure how
far this omission is relevant to the volition issue; one could just as
easily argue that works of genius or supernormal perception are as
involuntary as the products of pathology. However, it is another
example of how researchers tend to cherry pick phenomena that serve
their theoretical agendas, and ignore pieces that may prove
inconvenient.
The conflicting views on automatisms are interesting because
they reveal the ambiguities in much of the data, and how this can be
used to serve several contrasting worldviews. It's also sobering to
know how little progress has been made in this field the last century
because of scientific prejudice.
Chapter 6: Unusual experiences near death and related phenomena

[E.W. Kelly, B. Greyson & E.F. Kelly]
Chapter 6 primarily concerns Near Death Experiences, but also

covers Out of Body Experiences, Death Bed Visions, Veridical and
Collective Apparitions. Much of what's covered will be familiar to
those who've followed NDE debates over the years. The basic
phenomenology of NDEs is described, and conventional
neurophysiological explanations considered. The authors conclude that
whilst each of these explanations has some merit, none can explain the
NDE as a whole.
The authors then consider those aspects of the NDE that seem to
defy conventional models. These include enhanced mentation,
veridical Out-of-Body perceptions and visions of deceased
acquaintances. Enhanced mentation refers to the report that 'full-
fledged mentation, either normal or even enhanced mental activity (p.
386)' occurs when according to conventional theory, mental activity
should be receding. The writers acknowledge the difficulty of assessing
such subjective reports, but are impressed by their frequency. Veridical
OBE perceptions refer to the experiencer seeing or hearing things that
196
Colborn
they should have been inaccessible during the period of

unconsciousness.
Veridical impressions are also hard to verify, first because well-
attested cases are admittedly rare, or according to some authors,
probably non-existent (Blackmore, 1993). Second, it's hard to eliminate
the possibility that unconscious person might have seen or heard
something that was later incorporated into the NDE account, that
unwittingly lends it veridicality (Blackmore, 1993).
However, there are some aspects of some NDEs that do suggest
something strange is going on. Firstly, a number of researchers have
found that the memory of events happening just before or after
consciousness loss are often confused or absent, which contrasts with
the 'enhanced mentation' reports mentioned above. Second, there are
rare cases where events were recalled that were occurring when the
experiencer's EEG readings were flat. This is potentially very
problematic for conventional neuroscience theories of a 'global
workspace' because such theories say that normal mental functioning
requires synchronous EEG oscillations in the brain. Thinking seems to
be happening when these theories say it is impossible.
NDE research suffers because it's often popularly framed as a
dispute between those who 'believe' in the afterlife interpretation and
those who think that conventional scientific explanations are sufficient
(Kellehear, 1995). This is unfortunate, because such polarized thinking
might very well mask genuine anomalies that point to unknown mind/
brain functions. However, more and better well-attested cases are still
needed, even if they are rare and difficult to find.
Chapter 7: Genius [E.F. Kelly & M. Grosso]
In many senses the products of genius are the most problematic

for mainstream psychology, because they are poorly understood but
relied upon for scientific breakthroughs. Myers' view of genius was
that it 'should be regarded as a power of utilizing a wider range than
other men can utilize of faculties…innate in all…. (Human Personality.
Vol. 1, p. 71).' He spoke of a 'Subliminal uprush,' that is the
manifestation of strength and concentration of some inward unifying
control into the conscious mind. This 'uprush' is often achieved
through various forms of automatism and even hallucination.
197
This view is compatible with Hutchinson's (1939) description of

the stages of creativity (preparation, incubation, illumination and
verification). However, there is a modern school that that sees genius
as an incremental application of ordinary consciousness and denies it's
quantitatively different at all (Weisberg, 1999; Perkins, 2000). Kelly and
Grosso rightly comment that this is an especially egregious example of
a willful disregard of evidence that creativity is actually a very
mysterious and complex phenomenon. The incommensurability of
creative thought with conscious, logical thought is an important
example of where computational models fall down in describing the
creative process. Myers was right to highlight that many of the creative
aspects of genius lie beyond speech, in an almost mystical realm. The
authors also highlight the limits of associationism and neural network
views in this respect.
Kelly & Grosso are also right to raise the idea that genius is
somehow a personality in transformation, with strong links to the
mystical. They summarise four areas in which progress needs to be
made. Firstly as with other topics, an expansion of current psychology
to accommodate data is needed, second is to address the problem of
intentionality, third is to focus on the transformative nature of genius
and lastly recognition of the links between mysticism and genius.
Chapter 8: Mystical Experience [E.F. Kelly & M. Grosso].
This chapter is in some ways the heart of the book, because it

describes experiences that on occasion seem to show that 'pure,
unitary, undifferentiated, self-reflexive consciousness' can exist (p.573).
This prima facie contradicts functionalist and computationalist theories
of the human mind because it suggests that consciousness has a
'central and even supreme reality' over and above function.
The authors begin by providing an overview of the
phenomenology of mystical experience. These include ineffability (that
these experiences describe expression), that they seem to be in some
sense states of non or trans-verbal knowledge or insight, that they are
intense and not often sustained and that the experiences often exhibit
certitude as to its importance. To quote James (1902), p. 321); 'In mystic
states we both become one with the absolute and we become aware of
our oneness.'
198
Colborn
A key point is whether persons who have had one or more

mystical experiences are really talking about the same thing, or
whether they are really talking about distinct experiences that have
been moulded by their own cultures and only afterwards interpreted
as being similar. Constructivists such as Katz (1978) argued that since
any experience is filtered by culture, one cannot really talk about a
common state of consciousness found across all cultures. However,
separate mystical traditions in very different seem to be very similar
and constructivism cannot account for the commonality in the reports.
The problem of 'objective significance,' remains. The authors list
several circumstantial arguments that mystics do, in fact, make contact
with reality in novel ways. There is the persistent connection between
mystical experience and genius. There are its connections with
automatism, also associated with creativity and genius. There are the
apparent connections with supernormal phenomena. But while
certainly suggestive, none of these features provide direct evidence
that mystical states are anything other than a novel function of the
brain.
However, current neurobiological theories of mystical states are
not very good. The authors note that researchers commonly (1) fail to
come to grips with the full-blown phenomenology of the mystical
experience, (2) often have very poor supporting data and (3) are
excessively willing to spin out elaborate neurobiological just-so stories.
A case in point is the claimed association between mystical states and
Temporal Lobe Epilepsy (Saver & Rabin, 1997), for which the authors
report 'no credible evidence (p. 534).'
The latter part of the chapter concern the induction of mystical
experiences by drugs, including LSD and Ketamine. Again, the
associations of these states with other anomalous experiences like NDE
have been exaggerated, but there's enough common ground to see
these experiences as related but not identical. The authors also
comment on opportunities for further research, which are hampered
by the current legal restrictions on drugs. An important problem with
the assessment of mystical states is that mainstream neuroscience still
regards introspection as a dangerous way to gather evidence,
especially for non-ordinary states of consciousness. It's all too easy to
dismiss these accounts as not really showing what they purport they
show. However, the study of such experiences is important if only
because it forces us to confront an important dilemma for psychology,
199
which is; how do we make an objective science of something that can

only be directly accessed by subjective means?
Chapter 9: Toward a psychology for the 21st Century [E.F. Kelly]
The final chapter begins with a reassessment of Myers' theory

and assesses its implications for future psychology. It begins with a
brief look at contemporary (1900s) reviews of Human Personality,
which included concerns with the definition of 'Subliminal Mind' and
with the inclusion of 'rogue phenomena.'
Several issues already covered are also raised, notably Myers'
awareness that psychology must be expanded to accommodate
territory previously held to belong to religion or metaphysics.
Especially important is Myers' plea for a flexible approach to research,
and the need to adapt research methods to new situations rather than
trying to fit new problems into existing methods.
Kelly concludes that the current status of Myers' Subliminal self
is somewhat mixed. He thinks that the conception remains
counterintuitive and logically difficult, and that the supporting
evidence is 'less than compelling.' (p.594). However, Kelly regards it as
'definitely possible and perhaps even probable, especially in the light
of…mystical experiences' (p.595) that some sort of transmission theory
is correct. If this is so, then Survival after death, Myers' central concern,
becomes at least a possibility.
Can this view of the human psyche be reconciled with
contemporary science? One of the strengths of the theory is that it
connects, albeit in general terms, both 'normal' and 'supernormal'
phenomena. For example, it makes explicit the connection between psi
phenomena and dreaming, genius and mysticism. Kelly also claims
that the theory has some predictive value; Myers seems to have
anticipated NDEs, and his concept of a 'permeable' boundary between
the subliminal and normal consciousness predates the work of
Thalbourne (1998) and Hartmann (1991) by a century.
The authors of IM believe that the evidence they present is
'sufficient to falsify all forms of biological naturalism, the current
physicalist consensus on mind-brain relations. 'The mind is
“irreducible”…. There is at least a fundamental bifurcation in nature
that cannot be accounted for in [conventional physicalist terms] and we
200
Colborn
seem driven toward some sort of animist or pluralist alternative.'

(p.605).
Despite the cohorts of data presented, I remain uneasy with this
conclusion, for both evidential and conceptual reasons. First, even
allowing for the length of the book, there are some significant
omissions in the evidence. The relative neglect of experimental
parapsychological research is strange given that the authors think that
it's crucial in their arguments. In fairness, the psi debates are
referenced in an annotated bibliography of psychical research, but
more discussion of the results, and their interpretation in terms of
Myers' Subliminal Mind would have helped.
One consequence of making the data from psychical research
central to the arguments for the mind being 'irreducible' is the
importation of a number of practical and conceptual problems from
that field. Reading IM, one is sometimes left with the impression that
these 'rogue phenomena' are accessible to anyone who has eyes to see.
This is certainly not the case with psi results, which are often weak,
intermittent and context dependent (Hansen, 2001). Experimenter
effects are rampant in this field, and 'decline effects' may also be
significant (Colborn, 2007). The reports of many of the more outré
phenomena like 'maternal impressions' are remote in time and space,
and might also be declining (Stevenson, 1997). Any model of an
alternate picture of mind must convincingly accommodate these issues,
and it's not clear that this has been done.
There are also sociological problems to bridge. One needs also to
acknowledge that academe plays an active part in what Weber termed
'the disenchantment of the world' (Hansen, 2001). This concept shows
that the erasure of any magical notion of the mind is and probably will
remain, central to any future mainstream psychology. Many of the
phenomena presented in IM (Stigmata, Reincarnation, Mystical
experience) are just too closely associated with religious or mystical
views of the world to ever really be part of the rational academic
universe.
Another problem is data ambiguity. Anyone carefully reviewing
the data of psychical research will find that quite often the data can be
made to cut both ways (Bauer, 2001; Collins & Pinch, 1982), as
advocates and counter-advocates interpret the same data in different
ways according to their a priori expectations. This issue is not just
salient to the existence or otherwise of a phenomenon; quite often we
201
cannot be sure whether the data really points to dualism or not.

Laboratory psi phenomena might well indicate that some form of
dualism is necessary, but then again they might not. Godbey (1978)
argued fairly coherently that psi was not incompatible with central
state materialism. It is quite possible to build a coherent case for
dualism, and a coherent case against, using much of the same data.
Differences in interpretation often arise because of a difference in
a priori expectations about the world, and many of these are not
directly testable. Metaphysics is the 'elephant in the room' of
consciousness studies. As Midgley (2001) points out, many current
discussions about consciousness are really about clashing metaphysics.
This denial of metaphysics is also evident in IM. Whilst the
authors (I believe correctly) reject the simplistic assumptions of the
mainstream, they fail to make sufficiently explicit that often the
argument is not over empirical evidence but over ideological and
metaphysical difference. This is not to deny that the cumulative
evidence presented points to a far richer, more complex and interesting
model of mind than is currently advocated by the mainstream. But the
central issue of subjective experience (as outlined on pp. 24—29)
remains relatively untouched, and perhaps untouchable (see Nagel,
1974).
Finally, the authors examine two alternative models of mind,
which they consider congruent with the evidence presented. The first
is a non-Cartesian dualist-interactionist model, and the second a
neutral-monist model. I remain suspicious of dualist models for
reasons that have nothing to do with the 'denial' of rogue phenomena.
Dualism was not only abandoned because of prejudice; the Cartesian
split was quite artificial and arbitrary, introduced because Descartes
was too honest to ignore the strangeness of the subjective viewpoint,
but also unable to relate it to the mechanistic worldview he was
developing (Midgley, 2001). My own take is that dualism is in many
respects an 'observer effect' of the mind looking at itself and to take it
too literally is like looking in the mirror and mistaking the reflection
for a second person. In my view, it's fatal to lose sight of its artificial
nature. For this (metaphysical?) reason, I find the Kelly's suggestion of
a neutral-monist model of mind preferable to a return to a 'substance-
dualist' solution. As they rightly state, since quantum theory, 'mind'
and 'matter' have been redefined out of existence; neither are quite
202
Colborn
what we thought they were. Maybe 'promissory materialism' will also

eventually redefine itself out of existence.
The work of Henry Stapp (2005) seems especially promising here,
which seems to answer many of the criticisms of earlier Quantum
models and provide testable hypotheses of psychological phenomena
like attention. Stapp provides strong arguments to suggest that the
behaviour of low level brain constituents are saturated with quantum
effects, which puts the burden of proof on those who deny, not affirm
the relevance of quantum theory to brain science. These developments
do not mean the gap between theory and psychical research has been
fully bridged. Even physicists sympathetic to psi phenomena think that
current physics would need to be expanded to accommodate them
(Carr, 2008). This gap between theory and evidence is an important
obstacle to their case (Stent, 1972).
Despite these weaknesses, IM is an important work for a number
of reasons. Firstly, it rehabilitates the writings of Myers, which has
languished in obscurity for far too long. Second, it places 'psi' and
other 'rogue' phenomena in a wider psychological context, suggesting
ways that such findings might be integrated with mainstream
knowledge. Third, it provides a careful, thorough and systematic
critique of the weaknesses of mainstream claims. Fourth, it lays bear
the extent of the denial of pretty much every interesting psychological
phenomenon by mainstream science, mostly in the name of
demystification. For these reasons, it's a key work for
parapsychologists, and deserves also to be widely read by mainstream
cognitive scientists.
Publication Details: Kelly, E.F., Kelly, E.W., Crabtree, A., Gauld, A.,
Grosso, M. & Greyson, B. (2007) Irreducible Mind: Toward a psychology for
the 21st century. Rowman and Littlefield. 800 pp. ISBN-13:
978-0-7425-4792-6, ISBN-10: 0-7425-4792-2. Publication price: $79.95
(Cloth).
References
Bauer, H. (2001). Science or pseudoscience: Magnetic healing, psychic phenomena and

other heterodoxies. University of Illinois Press: Urbana and Chicago.
Blackmore, S.J. (1993). Dying to live. London: Grafton.
Buckman, R. (1996). What you really need to know about cancer. London: Pan
Macmillan.
203
Carr. B. (2008). Worlds Apart? Can Psychical Research Bridge the Gulf Between
Matter and Mind? Proceedings of the society for Psychical Research, 59, 1-96.
Chomsky, N. (1993). Language and Thought. Wakefield, RI: Moyer Bell.
Colborn, M.L.C. (2007). The decline effect in spontaneous and experimental
psychical research. Journal of the Society for Psychical Research, 71, 1-22.
Collins, H.M. & Pinch, T.J. (2005). Dr. Golem: How to think about medicine. Chicago:
University of Chicago Press.
Collins, H.M. & Pinch, T.J. (1982). Frames of meaning: the social construction of
extraordinary science. London: Routledge & Kegan Paul.
Coyne, J.C., Stefanek, M. and Palmer, S.C. (2007). Psychotherapy and survival in
cancer: The conflict between hope and evidence. Psychological Bulletin, 133,
367-394.
Dennett, D.C. (1991). Consciousness explained. Boston: Little, Brown.
Dennett, D.C. (1978). Towards a cognitive theory of consciousness. In Brainstorms:
Philosophical essays in mind and psychology. Montgomery: Bradford Books.
Feigl, H. (1958). The “mental” and the “physical.” In H.Feigl, M.Scriven, & G.
Maxwell (Eds.), Minnesota studies in the philosophy of science (Vol. 2, pp. 370-497).
Minneapolis: University of Minnesota Press.
Gauld, A. (1982). Mediumship and Survival. London: Heinemann.
Godbey, J.W. (1978). Central state materialism and parapsychology. In J. Ludwin
(Ed.) Philosophy and parapsychology. Prometheus books: Buffalo.
Hansen, G. P. (2001). The Trickster and the Paranormal. Xlibris.
Harpav, Y. (1997). Replicability of cognitive imaging of the cerebral cortex by PET and
fMRI: A survey of recent literature. http://human-brain.org/replicability.html,
accessed 31/07/08.
Hartmann, E. (1991). Boundaries in the mind: A new psychology of personality. New
York: Basic Books.
Hofstader, D.R. (2007). I am a strange loop. London: Basic Books.
Hutchinson E.D. (1939). Varieties of Insight in Humans. Psychiatry, 2, 323-332.
James, W. (1958). The varieties of religious experience. New York: Mentor. (Original
work published 1902).
James, W. (1890). The Principles of psychology (2 Vols). New York: Henry Holt.
Katz,S.T. (Ed.) (1978). Mysticism and philosophical analysis. New York: Oxford
University Press.
Kellehear, A. (1995). Review of Dying to Live: science and the Near-Death
Experience by Blackmore, S. (1993), Prometheus: Buffalo. Journal of Scientific
Exploration, 9, 312-14.
Kihlstrom, J.F. (1993) The psychological unconscious and the self. In G. R. Bock & J.
Marsh (Eds.), Experimental and theoretical studies of consciousness (pp. 147-156).
Chichester, UK: John Wiley & Sons.
Lipowski, Z.J. (1984). What does the word “psychosomatic” really mean? A
historical and semantic enquiry. Psychosomatic Medicine, 46, 153-171.
Meares, A. (1980). What can a cancer patient expect from intensive meditation?
Australian Family Physician, 9, 322-325.
204
Colborn
Meares, A. (1979). Meditation: A psychological approach to cancer treatment. The

Practitioner, 222, 119-122.
Meares, A. (1977). Atavistic regression as a factor in the remission of cancer.
Medical Journal of Australia, 2, 132-133.
Midgley, M. (2001). Science and Poetry. London: Routledge & Kegal Paul.
Moody, R.L. (1946). Bodily changes during abreaction. Lancet, 1, 934-935.
Moody, R.L. (1948). Bodily changes during abreaction. Lancet, 2, 964.
Myers, F.W.H. (1902). Human Personality and its survival of bodily death. (2 vols.).
London: Longmans Green.
Myers, F.W.H. (1885). Automatic writing – II. Proceedings of the Society for Psychical
Research, 3, 1-63.
Nagel, T. (1974). What Is it Like to Be a Bat? Philosophical Review, 83, 435-50.
O'Shea, M. (2008). Science made simple: the Brain (booklet distributed free with the
Independent newspaper in association with Oxford University Press).
Perkins, D.N. (2000). The Eureka Effect: The Art and logic of Breakthrough Thinking.
New York: W.W. Norton.
Pinker, S. (September 7th, 2004). Editorial, Newsweek.
Putnam, H. (1967). Psychological predicates. In W.H. Capitan & d.D. merrill (Eds.),
Art, Mind and Religion (pp. 37-48). Pittsburgh, PA: University of Pittsburgh
Press.
Rorty, R. (1979). Philosophy and the Mirror of Nature. Princeton, N.J.:Princeton
University Press.
Santayans, G. (1905). Life of Reason, Reason in Common Sense. Scribner's.
Saver, J.L. & Rabin, J. (1997). The neural substrates of religious experience. Journal
of Neuropsychiatry and Clinical Neuroscience, 9, 498-510.
Searle, J.R. (1992). The Rediscovery of Mind. Cambridge, MA: Harvard University
Press.
Searle, J. R. (1984). Minds, brains and science. Cambridge, MA: Harvard University
Press.
Searle, J.R. (1980). Minds, brains and programs. Behavioural and Brain sciences, 3,
417-424.
Stapp, H. (2005). Quantum interactive dualism: An alternative to materialism.
Journal of Consciousness Studies, 12, 43-58.
Stent, G. S. (1972). Prematurity and Uniqueness in Scientific Discovery. Scientific
American, December, 84-93.
Stevenson, I. (1997). Reincarnation and Biology: A contribution to the etiology of
birthmarks and birth defects. (2 vols.). New York: Praeger.
Thalbourne, M.A. (1998) Transliminality: Further correlates and a short measure.
Journal of the American Society for Psychical Research, 92, 402-419.
Turing, A. (1950). Computing machinery and intelligence. Mind, 59, 433-460.
Varela, F.J. (1997) Sleeping, dreaming and dying. Somerville, MA: Wisdom.
Weisberg, R.W. (1999). Creativity and Knowledge: A challenge to theories. In R.J.
Sternberg (Ed.), Handbook of Creativity (pp. 251-272). New York: Cambridge
University Press,
Wegner (2002). The illusion of conscious will. Cambridge, MA: MIT Press.
205

European Journal of Parapsychology v23-2 2008

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

European Journal of Parapsychology v23-2 2008

Uploaded by

Copyright:

Available Formats

European Journal of Parapsychology

Correlations between the EEGs of two spatially

Previous studies have reported correlations between the EEGs of

Since 1965, a series of studies have reported correlations between

stimulation on the EEG power of the non-stimulated subject B varied

Seventeen pairs of related subjects (friends, relatives, or partners; 12

As visual stimulus, a checkerboard pattern with 64 x 48 black and

EEG data were recorded from nineteen electrodes according to the

During the experiment, the subjects were seated in two spatially

Figure 2. Spatial arrangement of the experimental setup

Each subject was observed via a video-camera by one of the two

by Wackermann et al (2004). Figure 3 illustrates the procedure as a

Modification 1: Sampling for the bootstrap

In what follows, the previous sampling procedure turns out to

100,000 times. Fulfillment of the null hypothesis, meaning that sample

analyses revealed that with an increased number of subsets this range

Modification 2: Correction for degrees of freedom

A second concern refers to the collapsing of results over EEG

For the chi-square values obtained as study results, the

All analyses have been performed for two different time

Figure 7. Cumulative chi-square distributions from the Monte-Carlo simulations; A: 136 ms

Monte-Carlo simulations for both time windows were performed

Comparison with the previous method

In order to estimate the effect of the methodological

A further Monte-Carlo simulation was performed to investigate

The present study on correlations in the EEG of spatially

transmission of information by means of telepathy. Regardless of the

Dyadic EEG correlations re-examined:

Some years ago we were reporting results of an electrophysiologi-

Firstly, some introductory remarks for a reader who may not be

of the distribution of Q, empirical CDFs of Q were constructed from re-

1.0 1.0 1.0

0.8 0.8 0.8

Cumulative probability P(R )

Cumulative probability P(R )

0.4 0.4 0.4

0.2 0.2 0.2

0.0 0.0 0.0

1.0 1.0 1.0

0.8 0.8 0.8

Cumulative probability P(R )

Cumulative probability P(R )

0.4 0.4 0.4

0.2 0.2 0.2

0.0 0.0 0.0

Figure 1. Cumulative distribution functions of ranks R determined by numerical simulations

ranks, Z = Φ−1 ( R), would diminish with N1  N2 . It follows, then,

the simulations should not be applied too hastily. In the commented

perimental paradigm — that is, one yielding reproducible results across

Abbreviations and mathematical symbols

Lunar-magnetotail Encounters as Modulators of

Mind-matter interaction (MMI) effects are controversial partially

Mind-matter interaction (MMI) is the hypothesized ability of the

Geomagnetic effects on human behaviour

The influence of subtle electromagnetic effects on our human

problems, since we don’t know when MMI effects might be healing or

Lunar phase effects in MMI data

In an earlier study, Radin (1997) claimed evidence of MMI effects

The Fourmilab Retropsychokinesis Project

Etzold (2005) analysed data of the Fourmilab RetroPsychoKinesis

“retroPK“. RetroPK is a hypothesized ability to retroactively influence

The Moon-magnetotail interaction hypothesis

A hypothesis was suggested by earlier analyses that the Moon’s

While the effective parameter responsible for modulating MMI

ranks, Z = Φ−1 ( R), would diminish with N1 N2 . It follows, then,